[go: up one dir, main page]

CN105303179A - Fingerprint identification method and fingerprint identification device - Google Patents

Fingerprint identification method and fingerprint identification device Download PDF

Info

Publication number
CN105303179A
CN105303179A CN201510711591.8A CN201510711591A CN105303179A CN 105303179 A CN105303179 A CN 105303179A CN 201510711591 A CN201510711591 A CN 201510711591A CN 105303179 A CN105303179 A CN 105303179A
Authority
CN
China
Prior art keywords
fingerprint
sample image
training sample
recognition model
true
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201510711591.8A
Other languages
Chinese (zh)
Inventor
张涛
汪平仄
张胜凯
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Xiaomi Technology Co Ltd
Xiaomi Inc
Original Assignee
Xiaomi Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xiaomi Inc filed Critical Xiaomi Inc
Priority to CN201510711591.8A priority Critical patent/CN105303179A/en
Publication of CN105303179A publication Critical patent/CN105303179A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/12—Fingerprints or palmprints
    • G06V40/1382—Detecting the live character of the finger, i.e. distinguishing from a fake or cadaver finger
    • G06V40/1388—Detecting the live character of the finger, i.e. distinguishing from a fake or cadaver finger using image processing
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Human Computer Interaction (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)
  • Collating Specific Patterns (AREA)

Abstract

The invention relates to a fingerprint identification method and a fingerprint identification device. The method comprises the following steps: receiving a fingerprint image to be identified; and using a fingerprint identification model to identify the fingerprint to be identified, and determining whether the fingerprint to be identified is a true fingerprint, wherein the fingerprint identification model is obtained by using a fingerprint training sample set to train a convolution neural network in a classified manner. The fingerprint identification model built based on the convolution neural network can accurately extract all levels of features contained in the fingerprint image to be identified, namely, the precision of feature extraction is guaranteed, so that the result of fingerprint identification is more accurate and reliable.

Description

Fingerprint identification method, device
Technical field
The disclosure relates to fingerprint identification technology field, particularly relates to a kind of fingerprint identification method, device.
Background technology
At present, no matter fingerprint recognition is obtained for widespread use at civil area or military domain.Such as, for gate control system application, a lot of enterprise, unit have adopted fingerprint identification technology to carry out identity verify to user, to conduct interviews security control, and statistics work attendance etc.
The sorters such as at present, most fingerprint recognition system is all adopt traditional recognition methods, such as support vector machine (SupportVectorMachine is called for short SVM), carry out the true and false identification of fingerprint.In traditional recognition method, the extraction for fingerprint characteristic is chosen based on experience, and then carries out the learning training of such as SVM classifier according to the feature that these experiences are chosen, and utilizes the true and false identification training the sorter obtained to carry out fingerprint.
Summary of the invention
The disclosure provides a kind of fingerprint identification method, device, in order to improve the accuracy of the true and false identification of fingerprint.
According to the first aspect of disclosure embodiment, a kind of fingerprint identification method is provided, comprises:
Receive fingerprint image to be identified;
Adopt fingerprint recognition model to identify described fingerprint to be identified, determine whether described fingerprint to be identified is true fingerprint, described fingerprint recognition model obtains after adopting fingerprint training sample set pair convolutional neural networks to carry out classification based training.
Optionally, described method also comprises:
Described fingerprint image to be identified is normalized.
This technical scheme can comprise following beneficial effect: for the fingerprint image needing to identify that it is true and false, based on fingerprint training sample, the fingerprint recognition model that classification based training obtains is carried out to convolutional neural networks carry out true and false identification, to determine in this fingerprint image, whether fingerprint is true fingerprint by adopting.Due to the convolutional neural networks that learns based on the degree of depth can automatic learning to the detailed features information of the sandwich construction of sample fingerprint, thus ensure that the high precision of Finger print characteristic abstract in fingerprint image, make the result of fingerprint recognition more accurate.
Further, described method also comprises:
Obtain fingerprint training sample set, described fingerprint training sample is concentrated and is comprised true fingerprint training sample image and false fingerprint training sample image;
Described true fingerprint training sample image and described false fingerprint training sample image are input in convolutional neural networks, the characteristic coefficient between layer hidden node each in described convolutional neural networks are trained, obtains described fingerprint recognition model.
Optionally, described method also comprises:
Respectively described true fingerprint training sample image and described false fingerprint training sample image are normalized.
Optionally, described method also comprises:
Described true fingerprint training sample image is carried out to the mark of the first classification designator, described false fingerprint training sample image is carried out to the mark of the second classification designator.
Accordingly, described described true fingerprint training sample image and described false fingerprint training sample image are input in convolutional neural networks, characteristic coefficient between layer hidden node each in described convolutional neural networks is trained, obtains described fingerprint recognition model, comprising:
Current input training sample image is selected at random from described true fingerprint training sample image and described false fingerprint training sample image, be input in convolutional neural networks, characteristic coefficient between layer hidden node each in described convolutional neural networks is trained, obtains the output category label corresponding with current input training sample image;
If the difference between the first classification designator that output category label corresponding to current input training sample image is corresponding with current input training sample image or the second classification designator is greater than preset difference value, then adjust the characteristic coefficient between each layer hidden node of obtaining after described current input training sample image training.
Technique scheme can comprise following beneficial effect: by adopting the fingerprint training sample set including a large amount of true fingerprint training sample image and false fingerprint training sample image, learning classification training is carried out to convolutional neural networks, make the fingerprint recognition model that obtains can learn the multi-level characteristic information that comprises in each training sample image by automatic depth, thus improve and correctly identify based on this fingerprint recognition model the possibility that fingerprint image to be identified is true and false.
Further, described method also comprises:
Obtain fingerprint test sample book collection, described fingerprint test sample book is concentrated and is comprised true fingerprint test sample image and false fingerprint test sample image;
Respectively described fingerprint test sample book is concentrated to the discriminator result of described true fingerprint test sample image and described false fingerprint test sample image according to described fingerprint recognition model, determine the classification accuracy rate of described fingerprint recognition model.
Further, described method also comprises:
If described classification accuracy rate is less than predetermined threshold value, then iteration performs following process, until reach maximum iteration time or classification accuracy rate is greater than predetermined threshold value:
Upgrade described fingerprint training sample set;
Train according to the characteristic coefficient in fingerprint recognition model corresponding to fingerprint training sample set pair last iteration after upgrading between each layer hidden node, obtain fingerprint recognition model after renewal corresponding to current iteration;
Respectively described fingerprint test sample book is concentrated to the discriminator result of described survey true fingerprint test sample image and described false fingerprint test sample image according to fingerprint recognition model after the renewal that current iteration is corresponding, determine the classification accuracy rate of fingerprint recognition model after the renewal that current iteration is corresponding.
Further, the method also comprises:
Determine the maximum classification accuracy rate in the classification accuracy rate that each iteration is corresponding;
After determining the renewal corresponding with described maximum classification accuracy rate, fingerprint recognition model is target fingerprint model of cognition.
Technique scheme can comprise following beneficial effect: by the process of successive ignition training, test, can ensure that the target fingerprint model of cognition obtained has best accuracy.
According to the second aspect of disclosure embodiment, a kind of fingerprint identification device is provided, comprises:
Receiver module, is configured to receive fingerprint image to be identified;
Identification module, be configured to adopt fingerprint recognition model to identify described fingerprint to be identified, determine whether described fingerprint to be identified is true fingerprint, described fingerprint recognition model obtains after adopting fingerprint training sample set pair convolutional neural networks to carry out classification based training.
Optionally, described device also comprises:
First normalization module, is configured to be normalized described fingerprint image to be identified.
This technical scheme can comprise following beneficial effect: for the fingerprint image needing to identify that it is true and false, based on fingerprint training sample, the fingerprint recognition model that classification based training obtains is carried out to convolutional neural networks carry out true and false identification, to determine in this fingerprint image, whether fingerprint is true fingerprint by adopting.Due to the convolutional neural networks that learns based on the degree of depth can automatic learning to the detailed features information of the sandwich construction of sample fingerprint, thus ensure that the high precision of Finger print characteristic abstract in fingerprint image, make the result of fingerprint recognition more accurate.
Further, described device also comprises:
First acquisition module, is configured to obtain fingerprint training sample set, and described fingerprint training sample is concentrated and comprised true fingerprint training sample image and false fingerprint training sample image;
Training module, described true fingerprint training sample image and described false fingerprint training sample image is configured to be input in convolutional neural networks, characteristic coefficient between layer hidden node each in described convolutional neural networks is trained, obtains described fingerprint recognition model.
Optionally, described device also comprises:
Second normalization module, is configured to be normalized described true fingerprint training sample image and described false fingerprint training sample image respectively.
Optionally, described device also comprises:
Mark module, is configured to the mark described true fingerprint training sample image being carried out to the first classification designator, described false fingerprint training sample image is carried out to the mark of the second classification designator;
Described training module comprises:
Training submodule, be configured to from described true fingerprint training sample image and described false fingerprint training sample image, select current input training sample image at random, be input in convolutional neural networks, characteristic coefficient between layer hidden node each in described convolutional neural networks is trained, obtains the output category label corresponding with current input training sample image;
Adjustment submodule, when the difference be configured between the output category label that current input training sample image is corresponding first classification designator corresponding with current input training sample image or the second classification designator is greater than preset difference value, adjust the characteristic coefficient between each layer hidden node of obtaining after described current input training sample image training.
Technique scheme can comprise following beneficial effect: training module includes the fingerprint training sample set of a large amount of true fingerprint training sample image and false fingerprint training sample image by adopting, learning classification training is carried out to convolutional neural networks, make the fingerprint recognition model that obtains can learn the multi-level characteristic information that comprises in each training sample image by automatic depth, thus improve and correctly identify based on this fingerprint recognition model the possibility that fingerprint image to be identified is true and false.
Further, described device also comprises:
Second acquisition module, be configured to obtain fingerprint test sample book collection, described fingerprint test sample book is concentrated and is comprised true fingerprint test sample image and false fingerprint test sample image;
First determination module, be configured to the discriminator result respectively described fingerprint test sample book being concentrated to described true fingerprint test sample image and described false fingerprint test sample image according to described fingerprint recognition model, determine the classification accuracy rate of described fingerprint recognition model.
Optionally, described device also comprises:
Iteration execution module, is configured to when described classification accuracy rate is less than predetermined threshold value, and iteration performs following process, until reach maximum iteration time or classification accuracy rate is greater than predetermined threshold value:
Upgrade described fingerprint training sample set;
Train according to the characteristic coefficient in fingerprint recognition model corresponding to fingerprint training sample set pair last iteration after upgrading between each layer hidden node, obtain fingerprint recognition model after renewal corresponding to current iteration;
Respectively described fingerprint test sample book is concentrated to the discriminator result of described survey true fingerprint test sample image and described false fingerprint test sample image according to fingerprint recognition model after the renewal that current iteration is corresponding, determine the classification accuracy rate of fingerprint recognition model after the renewal that current iteration is corresponding.
Optionally, described device also comprises:
Second determination module, is configured to determine the maximum classification accuracy rate in the classification accuracy rate that each iteration is corresponding;
3rd determination module, after being configured to determine the renewal corresponding with described maximum classification accuracy rate, fingerprint recognition model is target fingerprint model of cognition.
Technique scheme can comprise following beneficial effect: by the process of successive ignition training, test, can ensure that the target fingerprint model of cognition obtained has best accuracy.
According to the third aspect of disclosure embodiment, a kind of fingerprint identification device is provided, comprises:
Processor;
Be configured to the storer of storage of processor executable instruction;
Wherein, described processor is configured to:
Receive fingerprint image to be identified;
Adopt fingerprint recognition model to identify described fingerprint to be identified, determine whether described fingerprint to be identified is true fingerprint, described fingerprint recognition model obtains after adopting fingerprint training sample set pair convolutional neural networks to carry out classification based training.
Should be understood that, it is only exemplary and explanatory that above general description and details hereinafter describe, and can not limit the disclosure.
Accompanying drawing explanation
Accompanying drawing to be herein merged in instructions and to form the part of this instructions, shows and meets embodiment of the present disclosure, and is used from instructions one and explains principle of the present disclosure.
Fig. 1 is the process flow diagram of a kind of fingerprint identification method embodiment one according to an exemplary embodiment;
Fig. 2 is the schematic network structure of convolutional neural networks;
Fig. 3 is the process flow diagram of a kind of fingerprint identification method embodiment two according to an exemplary embodiment;
Fig. 4 is the process flow diagram of a kind of fingerprint identification method embodiment three according to an exemplary embodiment;
Fig. 5 is the block diagram of a kind of fingerprint identification device embodiment one according to an exemplary embodiment;
Fig. 6 is the block diagram of a kind of fingerprint identification device embodiment two according to an exemplary embodiment;
Fig. 7 is the block diagram of a kind of fingerprint identification device embodiment three according to an exemplary embodiment;
Fig. 8 is the block diagram of a kind of fingerprint identification device according to an exemplary embodiment;
Fig. 9 is the block diagram of the another kind of fingerprint identification device according to an exemplary embodiment.
By above-mentioned accompanying drawing, illustrate the embodiment that the disclosure is clear and definite more detailed description will be had hereinafter.These accompanying drawings and text description be not in order to limited by any mode the disclosure design scope, but by reference to specific embodiment for those skilled in the art illustrate concept of the present disclosure.
Embodiment
Here will be described exemplary embodiment in detail, its sample table shows in the accompanying drawings.When description below relates to accompanying drawing, unless otherwise indicated, the same numbers in different accompanying drawing represents same or analogous key element.Embodiment described in following exemplary embodiment does not represent all embodiments consistent with the disclosure.On the contrary, they only with as in appended claims describe in detail, the example of apparatus and method that aspects more of the present disclosure are consistent.
Fig. 1 is the process flow diagram of a kind of fingerprint identification method embodiment one according to an exemplary embodiment, this fingerprint identification method can be performed by fingerprint identification device, and this fingerprint identification device is specifically as follows the application A PP that server corresponding to the intelligent terminals such as mobile phone terminal, panel computer, PC or server are installed.This fingerprint identification device can also for the application A PP that the intelligent terminals such as mobile phone terminal, panel computer, PC or intelligent terminal are installed.
As shown in Figure 1, this fingerprint identification method comprises the following steps:
In a step 101, fingerprint image to be identified is received.
Wherein, the quantity that fingerprint identification device receives this fingerprint image to be identified of user's input can be one, also can be multiple.For the situation of multiple fingerprint image to be identified, need to carry out identifying processing for each fingerprint image to be identified respectively.
In a step 102, adopt fingerprint recognition model to identify fingerprint to be identified, determine whether fingerprint to be identified is true fingerprint, fingerprint recognition model obtains after adopting fingerprint training sample set pair convolutional neural networks to carry out classification based training.
In the present embodiment, convolutional neural networks is adopted to build fingerprint recognition model.Convolutional neural networks is the one of artificial neural network, has become the study hotspot of current speech analysis and field of image recognition.Its weights shared network structure makes it more to be similar to biological neural network, reduces the complexity of network model, decreases the quantity of weights.It is more obvious that this advantage shows when the input of network is multidimensional image, makes image directly as the input of network, can avoid feature extraction complicated in tional identification algorithm and data reconstruction processes.
The network structure of convolutional neural networks as shown in Figure 2, convolutional neural networks is the neural network of a multilayer, comprise multiple convolutional layer, down-sampling layer, full articulamentum and an output layer, every layer is made up of multiple two dimensional surface, and each plane is made up of multiple independent neuron.Can be regarded as from a plane to the mapping of next plane and make convolution algorithm.In the present embodiment, suppose that the fingerprint recognition model obtained based on convolutional neural networks has N Rotating fields, each weight coefficient connected and convolution kernel between adjacent two layers hidden node, that is above-mentioned characteristic coefficient is determined by the training of fingerprint training sample set.
Because the fingerprint recognition model in the present embodiment is in order to identify that fingerprint to be identified is true fingerprint or false fingerprint, therefore, be understandable that, above-mentioned to convolutional neural networks train used fingerprint training sample concentrate both comprised some true sample fingerprint images, also comprise some false sample fingerprint images.Concrete training process describes in subsequent embodiment.
What deserves to be explained is, to the object of the identification of fingerprint image to be identified in the present embodiment, be to determine that this fingerprint image is true or false, that is, in the present embodiment, the Output rusults of fingerprint recognition model only has two kinds, one is true fingerprint, and one is false fingerprint.
These are different from traditional fingerprint recognition implication.In traditional sense, the object of fingerprint recognition is whose fingerprint to identify this fingerprint image.
In the present embodiment, why will identify fingerprint to be identified is true or false based on such actual conditions: someone is in some object, extract other people fingerprint, thus logged in other people computer based on the fingerprint extracted, open other people door etc., for other people interests bring serious threat.
Due to someone the false fingerprint extracted, and the difference that may exist between the true fingerprint of this people is very little, thus is difficult to distinguish based on traditional feature extraction and Classification and Identification mode.Therefore, present embodiments provide the convolutional neural networks based on degree of depth study, carry out the structure of fingerprint recognition model, make the extraction fineness of fingerprint characteristic higher more accurate.
In actual use, fingerprint image to be identified is input in above-mentioned fingerprint recognition model, through the feature extraction that fingerprint recognition model carries out successively to this fingerprint image to be identified, finally by the sorter of the last one deck of fingerprint recognition model, export the classification results of this fingerprint image to be identified, learn that the fingerprint corresponding to this fingerprint image to be identified is true fingerprint or false fingerprint according to this classification results.Suppose, in the recognition result that fingerprint recognition model exports and classification results, to represent true fingerprint with 1, represent false fingerprint with 2, if the recognition result of so above-mentioned fingerprint image to be identified is 1, illustrate that the fingerprint in this fingerprint image to be identified is true fingerprint.
Optionally, after the true and false property identifying fingerprint image to be identified, can mark this fingerprint image to be identified, if what such as this fingerprint image to be identified was corresponding is false fingerprint, then stamp a label for this fingerprint image to be identified, this label is 2, to represent that this fingerprint image to be identified is for false fingerprint.
In addition, optionally, in order to ensure that recognition result accurately and reliably, can the fingerprint image to be identified being input to fingerprint recognition model be normalized.Such as according to preset dimension requirement, carry out the normalized of size, size is all normalized to such as 224 pixel * 224 pixel sizes.For another example, coordinate centralization, x-shearing, convergent-divergent and rotation etc. are normalized.
In the present embodiment, adopt fingerprint training sample set pair convolutional neural networks to carry out classification based training, thus construct fingerprint recognition model, and then based on this fingerprint recognition model, fingerprint to be identified is identified, determine whether fingerprint to be identified is true fingerprint.The above-mentioned fingerprint recognition model built due to the convolutional neural networks that learns based on the degree of depth can automatic learning to the multilayer detailed features information of fingerprint image, thus ensure that the high-fineness of Finger print characteristic abstract in fingerprint image, make the result of fingerprint recognition more accurate.
Below in conjunction with embodiment illustrated in fig. 3, the formation process of above-mentioned fingerprint recognition model and the learning training process of convolutional neural networks are introduced.
Fig. 3 is the process flow diagram of a kind of fingerprint identification method embodiment two according to an exemplary embodiment, and as shown in Figure 3, the building process of fingerprint recognition model is as follows:
In step 201, obtain fingerprint training sample set, fingerprint training sample is concentrated and is comprised true fingerprint training sample image and false fingerprint training sample image.
In the present embodiment, in order to ensure fingerprint recognition model accurately and reliably, during training, need to gather a large amount of true fingerprint training sample image and false fingerprint training sample image, such as true fingerprint training sample image 2,000,000, false fingerprint training sample image 1,000,000.
In step 202., true fingerprint training sample image is carried out to the mark of the first classification designator, false fingerprint training sample image is carried out to the mark of the second classification designator.
In order to distinguish true fingerprint training sample image and false fingerprint training sample image, and it is whether accurate in order to determine to train the fingerprint recognition model obtained, respectively true fingerprint training sample image and false fingerprint training sample image are carried out to the mark of classification designator, suppose that true fingerprint training sample image is all labeled as the first classification designator 1, false fingerprint training sample image is all labeled as the second classification designator 2.
In addition, in order to minimize the impact of sample variation on fingerprint recognition model training process, in the present embodiment, also respectively true fingerprint training sample image and false fingerprint training sample image being normalized, ensureing the in the same size of each sample image.Such as size, coordinate centralization, x-shearing, convergent-divergent and rotation etc. are normalized.
And then, can at random the true fingerprint training sample image after normalized and false fingerprint training sample image be input in convolutional neural networks, characteristic coefficient between layer hidden node each in convolutional neural networks is trained, obtains fingerprint recognition model.
In step 203, current input training sample image is selected at random from true fingerprint training sample image and false fingerprint training sample image, be input in convolutional neural networks, characteristic coefficient between layer hidden node each in convolutional neural networks is trained, obtains the output category label corresponding with current input training sample image.
In step 204, determine whether the difference between the first classification designator that output category label corresponding to current input training sample image is corresponding with current input training sample image or the second classification designator is greater than preset difference value, if be greater than preset difference value, then perform step 205.Otherwise, upgrade current input training sample image, until all training sample image are all transfused to.
In step 205, the characteristic coefficient between each layer hidden node obtained after current input training sample image training is adjusted.
After step 205, upgrade current input training sample image, until all training sample image are all transfused to.
In above-mentioned training process, Stochastic choice training sample image is concentrated to be input to convolutional neural networks from fingerprint training sample, through classification learning training, output category result, i.e. above-mentioned output category label.
If the classification designator that this output category label is corresponding with this input training sample image (corresponding first classification designator of true fingerprint training sample image, corresponding second classification designator of false fingerprint training sample image) between difference be greater than preset difference value, illustrate and train the fingerprint recognition model obtained accurate not enough through this input training sample image, need to adjust it, after adjustment, need the input training then carrying out other training sample image follow-up.
Wherein, difference between the classification designator that output category label is corresponding with the training sample image of input is the value of the loss function that output layer exports, and can calculate according to such as Euclidean distance, mahalanobis distance, Chebyshev's distance, the equidistant metric form of COS distance.
Specifically, whether training classification results correctly can be measured by the loss function of output layer, be such as multiple European radial basis function (EuclideanRadialBasisFunction), each loss function calculates input vector and namely inputs distance between characteristic coefficient vector that vector corresponding to training sample image and corresponding training obtain.If this distance is greater than certain distance threshold value, then need adjustment characteristic coefficient and convolution kernel.
Optionally, in the present embodiment, in order to improve the treatment effeciency of training process further, reducing the number of times of above-mentioned adjustment, the mode of in batches training can be adopted.Specifically, all sample images concentrated by fingerprint training sample are random in batches, such as often criticize 100 sample images, wherein both comprised true fingerprint training sample image, also comprise false fingerprint training sample image.Random by each training sample image input in a collection of training sample image successively, obtain corresponding each output category label.After a collection of training sample image completes training, add up the ratio that training sample image quantity that the output category label of each training sample image in this batch of training sample image and the distance between corresponding classification designator be greater than predeterminable range accounts for this batch of training sample image sum, if this ratio is greater than certain threshold value, then adjust the characteristic coefficient between each layer hidden node in the fingerprint recognition model obtained through this batch of training sample image training.And then next group training sample image is input to premenstrual a collection of training sample image successively trains in the fingerprint recognition model obtained, until the training sample image of all batches has all been trained.Wherein, gradient descent method adjustment can be adopted to the adjustment of the characteristic coefficient between each layer hidden node.
In the present embodiment, by adopting the fingerprint training sample set including a large amount of true fingerprint training sample image and false fingerprint training sample image, learning classification training is carried out to convolutional neural networks, make the fingerprint recognition model that obtains can learn the multi-level characteristic information that comprises in each training sample image by automatic depth, thus improve and correctly identify based on this fingerprint recognition model the possibility that fingerprint image to be identified is true and false.
Fig. 4 is the process flow diagram of a kind of fingerprint identification method embodiment three according to an exemplary embodiment, as shown in Figure 4, after above-mentioned steps 205, also comprises the step of following test:
In step 301, obtain fingerprint test sample book collection, fingerprint test sample book is concentrated and is comprised true fingerprint test sample image and false fingerprint test sample image.
Wherein, the true fingerprint test sample image concentrated of this fingerprint test sample book is different from each training sample image that above-mentioned fingerprint training sample is concentrated with false fingerprint test sample image.
In addition, similar with training process, in order to distinguish each test sample image and the calculating for the ease of follow-up accuracy rate, respectively classification designator mark is carried out to true fingerprint test sample image and false fingerprint test sample image.Such as true fingerprint test sample image is all labeled as the first classification designator, false fingerprint test sample image is all labeled as the second classification designator.
In step 302, respectively fingerprint test sample book is concentrated to the discriminator result of true fingerprint test sample image and false fingerprint test sample image according to fingerprint recognition model, determine the classification accuracy rate of fingerprint recognition model.
In test process, random respectively each test sample image and true fingerprint test sample image and false fingerprint test sample image to be input in fingerprint recognition model, to obtain the classification designator exported of classifying.And then, according to the difference of the classification designator (i.e. the first classification designator or the second classification designator) of each output category label input corresponding with corresponding test sample image, determine the accuracy of fingerprint recognition model.
Specifically, according to predeterminable range difference metric modes such as such as Euclidean distance, mahalanobis distance, Chebyshev's distance, COS distance, calculate the distance difference between the classification designator of original tally corresponding to each test sample image and its classification designator exported through classifying respectively.And then each distance difference and predeterminable range threshold value that calculate acquisition can be compared, the distance quantity determining to be less than or equal to predeterminable range threshold value accounts for the ratio of the sample size that fingerprint test sample book collection comprises, and namely determines the accuracy rate of fingerprint recognition model.
If this accuracy rate is greater than necessarily higher accuracy rate threshold value, then illustrate that the accuracy of this fingerprint recognition model is good, be used for will obtaining good recognition effect in the follow-up true and false identification of fingerprint with this fingerprint recognition model.Contrary, if this accuracy rate is less than predetermined threshold value, illustrates and also need to carry out retraining to this fingerprint recognition model.
In step 303, determine whether classification accuracy rate is less than predetermined threshold value, if be less than predetermined threshold value, then iteration performs following process, until reach maximum iteration time or classification accuracy rate is greater than predetermined threshold value:
In step 304, described fingerprint training sample set is upgraded.
In step 305, train according to the characteristic coefficient in fingerprint recognition model corresponding to fingerprint training sample set pair last iteration after upgrading between each layer hidden node, obtain fingerprint recognition model after renewal corresponding to current iteration.
Within step 306, respectively fingerprint test sample book is concentrated to the discriminator result of true fingerprint test sample image and false fingerprint test sample image according to fingerprint recognition model after the renewal that current iteration is corresponding, determine the classification accuracy rate of fingerprint recognition model after the renewal that current iteration is corresponding.
In the present embodiment, when needing to carry out retraining to the fingerprint recognition model obtained based on above-mentioned fingerprint training sample set training, first need to upgrade fingerprint training sample set.Fingerprint training sample set after renewal is different from fingerprint training sample set before.
Above-mentionedly carry out the mode of training, testing based on the fingerprint training sample set pair fingerprint recognition model after upgrading, consistent with training before, test mode, do not repeat them here.
Be understandable that, if the accuracy rate of fingerprint recognition model is greater than predetermined threshold value after the renewal obtained with the fingerprint training sample set training after upgrading, then can terminate, with fingerprint recognition model after the renewal corresponding to the accuracy rate being greater than threshold value for target fingerprint model of cognition, in the true and false identification application of follow-up fingerprint.
But, if when reaching maximum iteration time, all there is no the fingerprint recognition model that accuracy rate is greater than predetermined threshold value, then, after iteration executes maximum iteration time, can following process be carried out:
In step 307, the maximum classification accuracy rate in the classification accuracy rate that each iteration is corresponding is determined.
In step 308, after determining the renewal corresponding with maximum classification accuracy rate, fingerprint recognition model is target fingerprint model of cognition.
That is, from the accuracy rate that each repetitive exercise, test are corresponding, determine maximum classification accuracy, and after finally determining the renewal that classification accuracy maximum with this is corresponding, fingerprint recognition model is target fingerprint model of cognition.
In above embodiment, by the process of successive ignition training, test, can ensure that the target fingerprint model of cognition obtained has best accuracy.
Fig. 5 is the block diagram of a kind of fingerprint identification device embodiment one according to an exemplary embodiment, and as shown in Figure 5, this device comprises: receiver module 11 and identification module 12.
Receiver module 11, is configured to receive fingerprint image to be identified.
Identification module 12, be configured to adopt fingerprint recognition model to identify described fingerprint to be identified, determine whether described fingerprint to be identified is true fingerprint, described fingerprint recognition model obtains after adopting fingerprint training sample set pair convolutional neural networks to carry out classification based training.
Wherein, the quantity that receiver module 11 receives this fingerprint image to be identified of user's input can be one, also can be multiple.For the situation of multiple fingerprint image to be identified, need to carry out identifying processing for each fingerprint image to be identified respectively.
In the present embodiment, convolutional neural networks is adopted to build fingerprint recognition model.Thus identification module 12 carries out by the fingerprint recognition model that training obtains the true and false identification inputting fingerprint image.
Convolutional neural networks is the one of artificial neural network, has become the study hotspot of current speech analysis and field of image recognition.Its weights shared network structure makes it more to be similar to biological neural network, reduces the complexity of network model, decreases the quantity of weights.It is more obvious that this advantage shows when the input of network is multidimensional image, makes image directly as the input of network, can avoid feature extraction complicated in tional identification algorithm and data reconstruction processes.
The network structure of convolutional neural networks as shown in Figure 2, convolutional neural networks is the neural network of a multilayer, comprise multiple convolutional layer, down-sampling layer, full articulamentum and an output layer, every layer is made up of multiple two dimensional surface, and each plane is made up of multiple independent neuron.Can be regarded as from a plane to the mapping of next plane and make convolution algorithm.In the present embodiment, suppose that the fingerprint recognition model obtained based on convolutional neural networks has N Rotating fields, each weight coefficient connected and convolution kernel between adjacent two layers hidden node, that is above-mentioned characteristic coefficient is determined by the training of fingerprint training sample set.
Because the fingerprint recognition model in the present embodiment is in order to identify that fingerprint to be identified is true fingerprint or false fingerprint, therefore, be understandable that, above-mentioned to convolutional neural networks train used fingerprint training sample concentrate both comprised some true sample fingerprint images, also comprise some false sample fingerprint images.Concrete training process describes in subsequent embodiment.
What deserves to be explained is, to the object of the identification of fingerprint image to be identified in the present embodiment, be to determine that this fingerprint image is true or false, that is, in the present embodiment, the Output rusults of fingerprint recognition model only has two kinds, one is true fingerprint, and one is false fingerprint.
These are different from traditional fingerprint recognition implication.In traditional sense, the object of fingerprint recognition is whose fingerprint to identify this fingerprint image.
In the present embodiment, why will identify fingerprint to be identified is true or false based on such actual conditions: someone is in some object, extract other people fingerprint, thus logged in other people computer based on the fingerprint extracted, open other people door etc., for other people interests bring serious threat.
Due to someone the false fingerprint extracted, and the difference that may exist between the true fingerprint of this people is very little, thus is difficult to distinguish based on traditional feature extraction and Classification and Identification mode.Therefore, present embodiments provide the convolutional neural networks based on degree of depth study, carry out the structure of fingerprint recognition model, make the extraction fineness of fingerprint characteristic higher more accurate.
In actual use, fingerprint image to be identified is input in above-mentioned fingerprint recognition model by identification module 12, through the feature extraction that fingerprint recognition model carries out successively to this fingerprint image to be identified, finally by the sorter of the last one deck of fingerprint recognition model, export the classification results of this fingerprint image to be identified, learn that the fingerprint corresponding to this fingerprint image to be identified is true fingerprint or false fingerprint according to this classification results.Suppose, in the recognition result that fingerprint recognition model exports and classification results, to represent true fingerprint with 1, represent false fingerprint with 2, if the recognition result of so above-mentioned fingerprint image to be identified is 1, illustrate that the fingerprint in this fingerprint image to be identified is true fingerprint.
Optionally, after identification module 12 identifies the true and false property of fingerprint image to be identified, identification module 12 can also mark this fingerprint image to be identified, if what such as this fingerprint image to be identified was corresponding is false fingerprint, then stamp a label for this fingerprint image to be identified, this label is 2, to represent that this fingerprint image to be identified is for false fingerprint.
Optionally, described device also comprises: the first normalization module 13.
First normalization module 13, is configured to be normalized described fingerprint image to be identified.
In order to ensure that recognition result accurately and reliably, can be normalized the fingerprint image to be identified being input to fingerprint recognition model by the first normalization module 13.Such as according to preset dimension requirement, carry out the normalized of size, size is all normalized to such as 224 pixel * 224 pixel sizes.For another example, coordinate centralization, x-shearing, convergent-divergent and rotation etc. are normalized.
In the present embodiment, adopt fingerprint training sample set pair convolutional neural networks to carry out classification based training, thus construct fingerprint recognition model, and then based on this fingerprint recognition model, fingerprint to be identified is identified, determine whether fingerprint to be identified is true fingerprint.The above-mentioned fingerprint recognition model built due to the convolutional neural networks that learns based on the degree of depth can automatic learning to the multilayer detailed features information of fingerprint image, thus ensure that the high-fineness of Finger print characteristic abstract in fingerprint image, make the result of fingerprint recognition more accurate.
Fig. 6 is the block diagram of a kind of fingerprint identification device embodiment two according to an exemplary embodiment, and as shown in Figure 6, on basis embodiment illustrated in fig. 5, described device also comprises: the first acquisition module 21 and training module 22.
First acquisition module 21, is configured to obtain fingerprint training sample set, and described fingerprint training sample is concentrated and comprised true fingerprint training sample image and false fingerprint training sample image.
In the present embodiment, in order to ensure fingerprint recognition model accurately and reliably, during training, the first acquisition module 21 needs to gather a large amount of true fingerprint training sample image and false fingerprint training sample image, such as true fingerprint training sample image 2,000,000, false fingerprint training sample image 1,000,000.
Training module 22, described true fingerprint training sample image and described false fingerprint training sample image is configured to be input in convolutional neural networks, characteristic coefficient between layer hidden node each in described convolutional neural networks is trained, obtains described fingerprint recognition model.
Optionally, described device also comprises: the second normalization module 23.
Second normalization module 23, is configured to be normalized described true fingerprint training sample image and described false fingerprint training sample image respectively.
In order to minimize the impact of sample variation on fingerprint recognition model training process, in the present embodiment, second normalization module 23 is normalized true fingerprint training sample image and false fingerprint training sample image respectively, ensures the in the same size of each sample image.Such as size, coordinate centralization, x-shearing, convergent-divergent and rotation etc. are normalized.
Optionally, described device also comprises: mark module 24.
Mark module 24, is configured to the mark described true fingerprint training sample image being carried out to the first classification designator, described false fingerprint training sample image is carried out to the mark of the second classification designator.
In order to distinguish true fingerprint training sample image and false fingerprint training sample image, and it is whether accurate in order to determine to train the fingerprint recognition model obtained, mark module 24 is adopted respectively true fingerprint training sample image and false fingerprint training sample image to be carried out to the mark of classification designator, suppose that true fingerprint training sample image is all labeled as the first classification designator 1, false fingerprint training sample image is all labeled as the second classification designator 2.
Specifically, described training module 22 comprises: training submodule 221 and adjustment submodule 222.
Training submodule 221, be configured to from described true fingerprint training sample image and described false fingerprint training sample image, select current input training sample image at random, be input in convolutional neural networks, characteristic coefficient between layer hidden node each in described convolutional neural networks is trained, obtains the output category label corresponding with current input training sample image.
Adjustment submodule 222, when the difference be configured between the output category label that current input training sample image is corresponding first classification designator corresponding with current input training sample image or the second classification designator is greater than preset difference value, adjust the characteristic coefficient between each layer hidden node of obtaining after described current input training sample image training.
In above-mentioned training process, training submodule 221 can concentrate Stochastic choice training sample image to be input to convolutional neural networks from fingerprint training sample, through classification learning training, and output category result, i.e. above-mentioned output category label.
If the classification designator that this output category label is corresponding with this input training sample image (corresponding first classification designator of true fingerprint training sample image, corresponding second classification designator of false fingerprint training sample image) between difference be greater than preset difference value, illustrate and train the fingerprint recognition model obtained accurate not enough through this input training sample image, adjustment submodule 222 needs to adjust it, needs the input training then carrying out other training sample image follow-up after adjustment.
Wherein, difference between the classification designator that output category label is corresponding with the training sample image of input is the value of the loss function that output layer exports, and can calculate according to such as Euclidean distance, mahalanobis distance, Chebyshev's distance, the equidistant metric form of COS distance.
Specifically, whether training classification results correctly can be measured by the loss function of output layer, be such as multiple European radial basis function (EuclideanRadialBasisFunction), each loss function calculates input vector and namely inputs distance between characteristic coefficient vector that vector corresponding to training sample image and corresponding training obtain.If adjustment submodule 222 determines that this distance is greater than certain distance threshold value, then need adjustment characteristic coefficient and convolution kernel.
Optionally, in the present embodiment, in order to improve the treatment effeciency of training process further, reducing the number of times of above-mentioned adjustment, the mode of in batches training can be adopted.Specifically, all sample images that fingerprint training sample can be concentrated by training submodule 221 are random in batches, such as often criticize 100 sample images, wherein both comprised true fingerprint training sample image, also comprise false fingerprint training sample image.Training submodule 221 and then random by each training sample image input in a collection of training sample image successively, obtains corresponding each output category label.After a collection of training sample image completes training, add up the ratio that training sample image quantity that the output category label of each training sample image in this batch of training sample image and the distance between corresponding classification designator be greater than predeterminable range accounts for this batch of training sample image sum, if this ratio is greater than certain threshold value, then adjusted the characteristic coefficient in the fingerprint recognition model obtained through this batch of training sample image training between each layer hidden node by adjustment submodule 222.Then training submodule 221 next group training sample image to be input to successively premenstrual a collection of training sample image trains in the fingerprint recognition model obtained, until the training sample image of all batches has all been trained again.Wherein, gradient descent method adjustment can be adopted to the adjustment of the characteristic coefficient between each layer hidden node.
In the present embodiment, by adopting the fingerprint training sample set including a large amount of true fingerprint training sample image and false fingerprint training sample image, learning classification training is carried out to convolutional neural networks, make the fingerprint recognition model that obtains can learn the multi-level characteristic information that comprises in each training sample image by automatic depth, thus improve and correctly identify based on this fingerprint recognition model the possibility that fingerprint image to be identified is true and false.
Fig. 7 is the block diagram of a kind of fingerprint identification device embodiment three according to an exemplary embodiment, and as shown in Figure 7, on the basis of above-described embodiment, described device also comprises: the second acquisition module 31 and the first determination module 32.
Second acquisition module 31, be configured to obtain fingerprint test sample book collection, described fingerprint test sample book is concentrated and is comprised true fingerprint test sample image and false fingerprint test sample image.
Wherein, the true fingerprint test sample image that this fingerprint test sample book that the second acquisition module 31 obtains is concentrated is different from each training sample image that above-mentioned fingerprint training sample is concentrated with false fingerprint test sample image.
First determination module 32, be configured to the discriminator result respectively described fingerprint test sample book being concentrated to described true fingerprint test sample image and described false fingerprint test sample image according to described fingerprint recognition model, determine the classification accuracy rate of described fingerprint recognition model.
In addition, similar with training process, in order to distinguish each test sample image and the calculating for the ease of follow-up accuracy rate, respectively classification designator mark is carried out to true fingerprint test sample image and false fingerprint test sample image.Such as true fingerprint test sample image is all labeled as the first classification designator, false fingerprint test sample image is all labeled as the second classification designator.
In test process, random each test sample image of being obtained by second acquisition module 31 and true fingerprint test sample image and false fingerprint test sample image are input in fingerprint recognition model respectively, obtain the classification designator exported of classifying.And then, the difference of the classification designator (i.e. the first classification designator or the second classification designator) of the input that each output category label that the first determination module 32 exports according to fingerprint recognition model is corresponding with corresponding test sample image, determines the accuracy rate of fingerprint recognition model.
Specifically, according to predeterminable range difference metric modes such as such as Euclidean distance, mahalanobis distance, Chebyshev's distance, COS distance, calculate the distance difference between the classification designator of original tally corresponding to each test sample image and its classification designator exported through classifying respectively.And then each distance difference and predeterminable range threshold value that calculate acquisition can be compared, the distance quantity determining to be less than or equal to predeterminable range threshold value accounts for the ratio of the sample size that fingerprint test sample book collection comprises, and namely determines the accuracy rate of fingerprint recognition model.
If this accuracy rate is greater than necessarily higher accuracy rate threshold value, then illustrate that the accuracy of this fingerprint recognition model is good, be used for will obtaining good recognition effect in the follow-up true and false identification of fingerprint with this fingerprint recognition model.Contrary, if this accuracy rate is less than predetermined threshold value, illustrates and also need to carry out retraining to this fingerprint recognition model.
Thus optionally, described device also comprises: iteration execution module 33.
Iteration execution module 33, is configured to when described classification accuracy rate is less than predetermined threshold value, and iteration performs following process, until reach maximum iteration time or classification accuracy rate is greater than predetermined threshold value:
Upgrade described fingerprint training sample set;
Train according to the characteristic coefficient in fingerprint recognition model corresponding to fingerprint training sample set pair last iteration after upgrading between each layer hidden node, obtain fingerprint recognition model after renewal corresponding to current iteration;
Respectively described fingerprint test sample book is concentrated to the discriminator result of described survey true fingerprint test sample image and described false fingerprint test sample image according to fingerprint recognition model after the renewal that current iteration is corresponding, determine the classification accuracy rate of fingerprint recognition model after the renewal that current iteration is corresponding.Optionally, described device also comprises: the second determination module 34 and the 3rd determination module 35.
Second determination module 34, is configured to determine the maximum classification accuracy rate in the classification accuracy rate that each iteration is corresponding.
3rd determination module 35, after being configured to determine the renewal corresponding with described maximum classification accuracy rate, fingerprint recognition model is target fingerprint model of cognition.
In the present embodiment, when needing to carry out retraining to the fingerprint recognition model obtained based on above-mentioned fingerprint training sample set training, first, iteration execution module 33 needs to upgrade fingerprint training sample set.Fingerprint training sample set after renewal is different from fingerprint training sample set before.
The mode of training, testing is carried out based on the fingerprint training sample set pair fingerprint recognition model after upgrading, consistent with training before, test mode, do not repeat them here.
Be understandable that, if the accuracy rate of fingerprint recognition model is greater than predetermined threshold value after the renewal obtained with the fingerprint training sample set training after upgrading, then can terminate, with fingerprint recognition model after the renewal corresponding to the accuracy rate being greater than threshold value for target fingerprint model of cognition, in the true and false identification application of follow-up fingerprint.
But, if when reaching maximum iteration time, all there is no the fingerprint recognition model that accuracy rate is greater than predetermined threshold value, then, after iteration executes maximum iteration time, can following process be carried out:
Second determination module 34 determines maximum classification accuracy from accuracy rate corresponding to each repetitive exercise, test, the 3rd determination module 35 so that after finally determining the renewal that classification accuracy maximum with this is corresponding fingerprint recognition model be target fingerprint model of cognition.
In above embodiment, by the process of successive ignition training, test, can ensure that the target fingerprint model of cognition obtained has best accuracy.
About the fingerprint identification device in above-described embodiment, wherein the concrete mode of modules, submodule, unit executable operations has been described in detail in about the embodiment of the method, will not elaborate explanation herein.
The foregoing describe built-in function and the structure of fingerprint identification device, as shown in Figure 8, in reality, this fingerprint identification device can be embodied as:
Processor;
Be configured to the storer of storage of processor executable instruction;
Wherein, described processor is configured to:
Receive fingerprint image to be identified;
Adopt fingerprint recognition model to identify described fingerprint to be identified, determine whether described fingerprint to be identified is true fingerprint, described fingerprint recognition model obtains after adopting fingerprint training sample set pair convolutional neural networks to carry out classification based training.
In above-described embodiment, fingerprint identification device adopts fingerprint training sample set pair convolutional neural networks to carry out classification based training, thus construct fingerprint recognition model, and then based on this fingerprint recognition model, fingerprint to be identified is identified, determine whether fingerprint to be identified is true fingerprint.The above-mentioned fingerprint recognition model built due to the convolutional neural networks that learns based on the degree of depth can automatic learning to the multilayer detailed features information of fingerprint image, thus ensure that the high-fineness of Finger print characteristic abstract in fingerprint image, make the result of fingerprint recognition more accurate.
Fig. 9 is the block diagram of the another kind of fingerprint identification device according to an exemplary embodiment.Such as, this fingerprint identification device 1900 may be provided in a server.With reference to Fig. 9, device 1900 comprises processing components 1922, and it comprises one or more processor further, and the memory resource representated by storer 1932, can such as, by the instruction of the execution of processing components 1922, application program for storing.The application program stored in storer 1932 can comprise each module corresponding to one group of instruction one or more.In addition, processing components 1922 is configured to perform instruction, to perform the fingerprint identification method provided in the various embodiments described above, comprising:
Receive fingerprint image to be identified;
Adopt fingerprint recognition model to identify described fingerprint to be identified, determine whether described fingerprint to be identified is true fingerprint, described fingerprint recognition model obtains after adopting fingerprint training sample set pair convolutional neural networks to carry out classification based training.
Device 1900 can also comprise the power management that a power supply module 1926 is configured to actuating unit 1900, and a wired or wireless network interface 1950 is configured to device 1900 to be connected to network, and input and output (I/O) interface 1958.Device 1900 can operate the operating system based on being stored in storer 1932, such as WindowsServerTM, MacOSXTM, UnixTM, LinuxTM, FreeBSDTM or similar.
Those skilled in the art, at consideration instructions and after putting into practice invention disclosed herein, will easily expect other embodiment of the present disclosure.The application is intended to contain any modification of the present disclosure, purposes or adaptations, and these modification, purposes or adaptations are followed general principle of the present disclosure and comprised the undocumented common practise in the art of the disclosure or conventional techniques means.Instructions and embodiment are only regarded as exemplary, and true scope of the present disclosure and spirit are pointed out by claim below.
Should be understood that, the disclosure is not limited to precision architecture described above and illustrated in the accompanying drawings, and can carry out various amendment and change not departing from its scope.The scope of the present disclosure is only limited by appended claim.

Claims (19)

1. a fingerprint identification method, is characterized in that, described method comprises:
Receive fingerprint image to be identified;
Adopt fingerprint recognition model to identify described fingerprint to be identified, determine whether described fingerprint to be identified is true fingerprint, described fingerprint recognition model obtains after adopting fingerprint training sample set pair convolutional neural networks to carry out classification based training.
2. method according to claim 1, is characterized in that, described method also comprises:
Obtain fingerprint training sample set, described fingerprint training sample is concentrated and is comprised true fingerprint training sample image and false fingerprint training sample image;
Described true fingerprint training sample image and described false fingerprint training sample image are input in convolutional neural networks, the characteristic coefficient between layer hidden node each in described convolutional neural networks are trained, obtains described fingerprint recognition model.
3. method according to claim 2, is characterized in that, described method also comprises:
Described true fingerprint training sample image is carried out to the mark of the first classification designator, described false fingerprint training sample image is carried out to the mark of the second classification designator.
4. method according to claim 3, it is characterized in that, described described true fingerprint training sample image and described false fingerprint training sample image are input in convolutional neural networks, characteristic coefficient between layer hidden node each in described convolutional neural networks is trained, obtain described fingerprint recognition model, comprising:
Current input training sample image is selected at random from described true fingerprint training sample image and described false fingerprint training sample image, be input in convolutional neural networks, characteristic coefficient between layer hidden node each in described convolutional neural networks is trained, obtains the output category label corresponding with current input training sample image;
If the difference between the first classification designator that output category label corresponding to current input training sample image is corresponding with current input training sample image or the second classification designator is greater than preset difference value, then adjust the characteristic coefficient between each layer hidden node of obtaining after described current input training sample image training.
5. method according to claim 2, is characterized in that, described method also comprises:
Obtain fingerprint test sample book collection, described fingerprint test sample book is concentrated and is comprised true fingerprint test sample image and false fingerprint test sample image;
Respectively described fingerprint test sample book is concentrated to the discriminator result of described true fingerprint test sample image and described false fingerprint test sample image according to described fingerprint recognition model, determine the classification accuracy rate of described fingerprint recognition model.
6. method according to claim 5, is characterized in that, the classification results corresponding according to each test sample image described, after determining the classification accuracy rate of described fingerprint recognition model, also comprises:
If described classification accuracy rate is less than predetermined threshold value, then iteration performs following process, until reach maximum iteration time or classification accuracy rate is greater than predetermined threshold value:
Upgrade described fingerprint training sample set;
Train according to the characteristic coefficient in fingerprint recognition model corresponding to fingerprint training sample set pair last iteration after upgrading between each layer hidden node, obtain fingerprint recognition model after renewal corresponding to current iteration;
Respectively described fingerprint test sample book is concentrated to the discriminator result of described survey true fingerprint test sample image and described false fingerprint test sample image according to fingerprint recognition model after the renewal that current iteration is corresponding, determine the classification accuracy rate of fingerprint recognition model after the renewal that current iteration is corresponding.
7. method according to claim 6, is characterized in that, also comprises:
Determine the maximum classification accuracy rate in the classification accuracy rate that each iteration is corresponding;
After determining the renewal corresponding with described maximum classification accuracy rate, fingerprint recognition model is target fingerprint model of cognition.
8. method according to claim 1, is characterized in that, described method also comprises:
Described fingerprint image to be identified is normalized.
9. method according to claim 2, is characterized in that, described method also comprises:
Respectively described true fingerprint training sample image and described false fingerprint training sample image are normalized.
10. a fingerprint identification device, is characterized in that, described device comprises:
Receiver module, is configured to receive fingerprint image to be identified;
Identification module, be configured to adopt fingerprint recognition model to identify described fingerprint to be identified, determine whether described fingerprint to be identified is true fingerprint, described fingerprint recognition model obtains after adopting fingerprint training sample set pair convolutional neural networks to carry out classification based training.
11. devices according to claim 10, is characterized in that, described device also comprises:
First acquisition module, is configured to obtain fingerprint training sample set, and described fingerprint training sample is concentrated and comprised true fingerprint training sample image and false fingerprint training sample image;
Training module, described true fingerprint training sample image and described false fingerprint training sample image is configured to be input in convolutional neural networks, characteristic coefficient between layer hidden node each in described convolutional neural networks is trained, obtains described fingerprint recognition model.
12. devices according to claim 11, is characterized in that, described device also comprises:
Mark module, is configured to the mark described true fingerprint training sample image being carried out to the first classification designator, described false fingerprint training sample image is carried out to the mark of the second classification designator.
13. devices according to claim 12, is characterized in that, described training module comprises:
Training submodule, be configured to from described true fingerprint training sample image and described false fingerprint training sample image, select current input training sample image at random, be input in convolutional neural networks, characteristic coefficient between layer hidden node each in described convolutional neural networks is trained, obtains the output category label corresponding with current input training sample image;
Adjustment submodule, when the difference be configured between the output category label that current input training sample image is corresponding first classification designator corresponding with current input training sample image or the second classification designator is greater than preset difference value, adjust the characteristic coefficient between each layer hidden node of obtaining after described current input training sample image training.
14. devices according to claim 11, is characterized in that, described device also comprises:
Second acquisition module, be configured to obtain fingerprint test sample book collection, described fingerprint test sample book is concentrated and is comprised true fingerprint test sample image and false fingerprint test sample image;
First determination module, be configured to the discriminator result respectively described fingerprint test sample book being concentrated to described true fingerprint test sample image and described false fingerprint test sample image according to described fingerprint recognition model, determine the classification accuracy rate of described fingerprint recognition model.
15. devices according to claim 14, is characterized in that, described device also comprises:
Iteration execution module, is configured to when described classification accuracy rate is less than predetermined threshold value, and iteration performs following process, until reach maximum iteration time or classification accuracy rate is greater than predetermined threshold value:
Upgrade described fingerprint training sample set;
Train according to the characteristic coefficient in fingerprint recognition model corresponding to fingerprint training sample set pair last iteration after upgrading between each layer hidden node, obtain fingerprint recognition model after renewal corresponding to current iteration;
Respectively described fingerprint test sample book is concentrated to the discriminator result of described survey true fingerprint test sample image and described false fingerprint test sample image according to fingerprint recognition model after the renewal that current iteration is corresponding, determine the classification accuracy rate of fingerprint recognition model after the renewal that current iteration is corresponding.
16. devices according to claim 15, is characterized in that, described device also comprises:
Second determination module, is configured to determine the maximum classification accuracy rate in the classification accuracy rate that each iteration is corresponding;
3rd determination module, after being configured to determine the renewal corresponding with described maximum classification accuracy rate, fingerprint recognition model is target fingerprint model of cognition.
17. devices according to claim 10, it is characterized in that, described device also comprises:
First normalization module, is configured to be normalized described fingerprint image to be identified.
18. devices according to claim 11, is characterized in that, described device also comprises:
Second normalization module, is configured to be normalized described true fingerprint training sample image and described false fingerprint training sample image respectively.
19. 1 kinds of fingerprint identification devices, is characterized in that, comprising:
Processor;
Be configured to the storer of storage of processor executable instruction;
Wherein, described processor is configured to:
Receive fingerprint image to be identified;
Adopt fingerprint recognition model to identify described fingerprint to be identified, determine whether described fingerprint to be identified is true fingerprint, described fingerprint recognition model obtains after adopting fingerprint training sample set pair convolutional neural networks to carry out classification based training.
CN201510711591.8A 2015-10-28 2015-10-28 Fingerprint identification method and fingerprint identification device Pending CN105303179A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510711591.8A CN105303179A (en) 2015-10-28 2015-10-28 Fingerprint identification method and fingerprint identification device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510711591.8A CN105303179A (en) 2015-10-28 2015-10-28 Fingerprint identification method and fingerprint identification device

Publications (1)

Publication Number Publication Date
CN105303179A true CN105303179A (en) 2016-02-03

Family

ID=55200423

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510711591.8A Pending CN105303179A (en) 2015-10-28 2015-10-28 Fingerprint identification method and fingerprint identification device

Country Status (1)

Country Link
CN (1) CN105303179A (en)

Cited By (28)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106056095A (en) * 2016-06-13 2016-10-26 北京小米移动软件有限公司 Fingerprint processing method and device
CN106096510A (en) * 2016-05-31 2016-11-09 北京小米移动软件有限公司 The method and apparatus of fingerprint recognition
CN106203298A (en) * 2016-06-30 2016-12-07 北京集创北方科技股份有限公司 Biological feather recognition method and device
CN106778820A (en) * 2016-11-25 2017-05-31 北京小米移动软件有限公司 Identification model determines method and device
WO2017161756A1 (en) * 2016-03-23 2017-09-28 乐视控股（北京）有限公司 Video identification method and system
CN107958217A (en) * 2017-11-28 2018-04-24 广州麦仑信息科技有限公司 A kind of fingerprint classification identifying system and method based on deep learning
CN108304859A (en) * 2017-12-29 2018-07-20 达闼科技（北京）有限公司 Image-recognizing method and cloud system
CN108520225A (en) * 2018-03-30 2018-09-11 南京信息工程大学 A Fingerprint Detection and Classification Method Based on Spatial Transformation Convolutional Neural Network
CN108564040A (en) * 2018-04-17 2018-09-21 杭州景联文科技有限公司 A kind of fingerprint activity test method based on depth convolution feature
CN109145834A (en) * 2018-08-27 2019-01-04 河南丰泰光电科技有限公司 A kind of fingerprint recognition neural network based and verification method
CN109214376A (en) * 2018-11-22 2019-01-15 济南浪潮高新科技投资发展有限公司 A kind of fingerprint identification method and device based on depth stratification
CN109508692A (en) * 2018-11-30 2019-03-22 深圳大学 A kind of gender identification method and system based on 3D fingerprint image
CN109870698A (en) * 2019-01-15 2019-06-11 北京百度网讯科技有限公司 An ultrasonic array obstacle detection result processing method and system
CN109949284A (en) * 2019-03-12 2019-06-28 天津瑟威兰斯科技有限公司 Deep learning convolution neural network-based algae cell analysis method and system
CN110059649A (en) * 2019-04-24 2019-07-26 济南浪潮高新科技投资发展有限公司 Level complementation convolutional neural networks model, robust fingerprint recognition methods and system
CN110326001A (en) * 2016-12-08 2019-10-11 维里迪乌姆Ip有限责任公司 The system and method for executing the user authentication based on fingerprint using the image captured using mobile device
CN110728968A (en) * 2019-10-14 2020-01-24 腾讯音乐娱乐科技（深圳）有限公司 Audio accompaniment information evaluation method and device and storage medium
CN110751105A (en) * 2019-10-22 2020-02-04 珠海格力电器股份有限公司 Finger image acquisition method and device and storage medium
CN111177585A (en) * 2018-11-13 2020-05-19 北京四维图新科技股份有限公司 Map POI feedback method and device
CN111435437A (en) * 2019-12-26 2020-07-21 珠海大横琴科技发展有限公司 PCB pedestrian re-recognition model training method and PCB pedestrian re-recognition method
CN111566659A (en) * 2018-12-13 2020-08-21 深圳市汇顶科技股份有限公司 Fingerprint identification device, method and electronic device
CN112437926A (en) * 2019-06-18 2021-03-02 神经技术Uab公司 Fast robust friction ridge imprint detail extraction using feed-forward convolutional neural networks
CN112966610A (en) * 2021-03-09 2021-06-15 金陵科技学院 Elevator intelligence fingerprint identification system based on increment study
CN113243018A (en) * 2020-08-01 2021-08-10 商汤国际私人有限公司 Target object identification method and device
CN113743371A (en) * 2021-09-22 2021-12-03 京东方科技集团股份有限公司 Fingerprint identification method and fingerprint identification device
CN113920550A (en) * 2021-10-29 2022-01-11 敦泰电子（深圳）有限公司 Finger authenticity identification method, electronic device and storage medium
CN114022911A (en) * 2021-10-12 2022-02-08 北京集创北方科技股份有限公司 Fingerprint identification method and device, electronic equipment and storage medium
US11875598B2 (en) 2019-07-17 2024-01-16 Huawei Technologies Co., Ltd. Fingerprint anti-counterfeiting method and electronic device

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103116744A (en) * 2013-02-05 2013-05-22 浙江工业大学 Fake fingerprint detection method based on markov random field (MRF) and support vector machine-k nearest neighbor (SVM-KNN) classification
CN103279744A (en) * 2013-05-28 2013-09-04 中国科学院自动化研究所 Multi-scale tri-mode texture feature-based method and system for detecting counterfeit fingerprints
CN103324944A (en) * 2013-06-26 2013-09-25 电子科技大学 Fake fingerprint detecting method based on SVM and sparse representation

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103116744A (en) * 2013-02-05 2013-05-22 浙江工业大学 Fake fingerprint detection method based on markov random field (MRF) and support vector machine-k nearest neighbor (SVM-KNN) classification
CN103279744A (en) * 2013-05-28 2013-09-04 中国科学院自动化研究所 Multi-scale tri-mode texture feature-based method and system for detecting counterfeit fingerprints
CN103324944A (en) * 2013-06-26 2013-09-25 电子科技大学 Fake fingerprint detecting method based on SVM and sparse representation

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
CHENGGANG WANG，等: "A DCNN Based Fingerprint Liveness Detection Algorithm with Voting Strategy", 《BIOMETRIC RECOGNITION》 *
LECUN Y，等: "Efficient BackProp", 《NEURAL NETWORKS: TRICKS OF THE TRADE》 *
NOGUEIRA R F, 等: "Evaluating software-based fingerprint liveness detection using convolutional networks and local binary patterns", 《BIOMETRIC MEASUREMENTS AND SYSTEMS FOR SECURITY AND MEDICAL APPLICATIONS (BIOMS) PROCEEDINGS》 *
YANG J,等: "Learn convolutional neural network for face anti-spoofing", 《ARXIV PREPRINT ARXIV》 *
杨远: "基于神经网络的目标图像识别方法研究", 《万方数据库》 *

Cited By (42)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017161756A1 (en) * 2016-03-23 2017-09-28 乐视控股（北京）有限公司 Video identification method and system
CN106096510A (en) * 2016-05-31 2016-11-09 北京小米移动软件有限公司 The method and apparatus of fingerprint recognition
CN106096510B (en) * 2016-05-31 2019-08-30 北京小米移动软件有限公司 Method and device for fingerprint recognition
CN106056095A (en) * 2016-06-13 2016-10-26 北京小米移动软件有限公司 Fingerprint processing method and device
CN106203298A (en) * 2016-06-30 2016-12-07 北京集创北方科技股份有限公司 Biological feather recognition method and device
CN106778820B (en) * 2016-11-25 2020-06-19 北京小米移动软件有限公司 Identification model determination method and device
CN106778820A (en) * 2016-11-25 2017-05-31 北京小米移动软件有限公司 Identification model determines method and device
CN110326001A (en) * 2016-12-08 2019-10-11 维里迪乌姆Ip有限责任公司 The system and method for executing the user authentication based on fingerprint using the image captured using mobile device
CN107958217A (en) * 2017-11-28 2018-04-24 广州麦仑信息科技有限公司 A kind of fingerprint classification identifying system and method based on deep learning
CN108304859A (en) * 2017-12-29 2018-07-20 达闼科技（北京）有限公司 Image-recognizing method and cloud system
CN108520225A (en) * 2018-03-30 2018-09-11 南京信息工程大学 A Fingerprint Detection and Classification Method Based on Spatial Transformation Convolutional Neural Network
CN108520225B (en) * 2018-03-30 2021-07-27 南京信息工程大学 A Fingerprint Detection and Classification Method Based on Spatial Transform Convolutional Neural Network
CN108564040A (en) * 2018-04-17 2018-09-21 杭州景联文科技有限公司 A kind of fingerprint activity test method based on depth convolution feature
CN109145834A (en) * 2018-08-27 2019-01-04 河南丰泰光电科技有限公司 A kind of fingerprint recognition neural network based and verification method
CN111177585A (en) * 2018-11-13 2020-05-19 北京四维图新科技股份有限公司 Map POI feedback method and device
CN109214376A (en) * 2018-11-22 2019-01-15 济南浪潮高新科技投资发展有限公司 A kind of fingerprint identification method and device based on depth stratification
WO2020107922A1 (en) * 2018-11-30 2020-06-04 深圳大学 3d fingerprint image-based gender recognition method and system
CN109508692B (en) * 2018-11-30 2020-06-16 深圳大学 A method and system for gender recognition based on 3D fingerprint images
CN109508692A (en) * 2018-11-30 2019-03-22 深圳大学 A kind of gender identification method and system based on 3D fingerprint image
CN111566659A (en) * 2018-12-13 2020-08-21 深圳市汇顶科技股份有限公司 Fingerprint identification device, method and electronic device
US11423690B2 (en) 2018-12-13 2022-08-23 Shenzhen GOODIX Technology Co., Ltd. Fingerprint identification apparatus and electronic device
US11475706B2 (en) 2018-12-13 2022-10-18 Shenzhen GOODIX Technology Co., Ltd. Fingerprint identification device, method and electronic device
CN111566659B (en) * 2018-12-13 2021-08-13 深圳市汇顶科技股份有限公司 Fingerprint identification device, method and electronic device
US11048910B2 (en) 2018-12-13 2021-06-29 Shenzhen GOODIX Technology Co., Ltd. Fingerprint identification device, method and electronic device
US11933921B2 (en) 2019-01-15 2024-03-19 Beijing Baidu Netcom Science And Technology Co., Ltd. Method and system for processing obstacle detection result of ultrasonic sensor array
CN109870698A (en) * 2019-01-15 2019-06-11 北京百度网讯科技有限公司 An ultrasonic array obstacle detection result processing method and system
CN109870698B (en) * 2019-01-15 2021-12-24 阿波罗智能技术(北京)有限公司 Ultrasonic array obstacle detection result processing method and system
CN109949284A (en) * 2019-03-12 2019-06-28 天津瑟威兰斯科技有限公司 Deep learning convolution neural network-based algae cell analysis method and system
CN110059649A (en) * 2019-04-24 2019-07-26 济南浪潮高新科技投资发展有限公司 Level complementation convolutional neural networks model, robust fingerprint recognition methods and system
CN112437926A (en) * 2019-06-18 2021-03-02 神经技术Uab公司 Fast robust friction ridge imprint detail extraction using feed-forward convolutional neural networks
CN112437926B (en) * 2019-06-18 2024-05-31 神经技术Uab公司 Fast and robust friction ridge imprint detail extraction using feed-forward convolutional neural network
US11875598B2 (en) 2019-07-17 2024-01-16 Huawei Technologies Co., Ltd. Fingerprint anti-counterfeiting method and electronic device
CN110728968A (en) * 2019-10-14 2020-01-24 腾讯音乐娱乐科技（深圳）有限公司 Audio accompaniment information evaluation method and device and storage medium
CN110751105B (en) * 2019-10-22 2022-04-08 珠海格力电器股份有限公司 Finger image acquisition method and device and storage medium
CN110751105A (en) * 2019-10-22 2020-02-04 珠海格力电器股份有限公司 Finger image acquisition method and device and storage medium
CN111435437A (en) * 2019-12-26 2020-07-21 珠海大横琴科技发展有限公司 PCB pedestrian re-recognition model training method and PCB pedestrian re-recognition method
CN113243018A (en) * 2020-08-01 2021-08-10 商汤国际私人有限公司 Target object identification method and device
CN112966610A (en) * 2021-03-09 2021-06-15 金陵科技学院 Elevator intelligence fingerprint identification system based on increment study
CN113743371A (en) * 2021-09-22 2021-12-03 京东方科技集团股份有限公司 Fingerprint identification method and fingerprint identification device
CN113743371B (en) * 2021-09-22 2024-07-19 京东方科技集团股份有限公司 Fingerprint identification method and fingerprint identification device
CN114022911A (en) * 2021-10-12 2022-02-08 北京集创北方科技股份有限公司 Fingerprint identification method and device, electronic equipment and storage medium
CN113920550A (en) * 2021-10-29 2022-01-11 敦泰电子（深圳）有限公司 Finger authenticity identification method, electronic device and storage medium

Similar Documents

Publication Publication Date Title
CN105303179A (en) Fingerprint identification method and fingerprint identification device
CN110348441B (en) Value-added tax invoice identification method and device, computer equipment and storage medium
CN112989035B (en) Method, device and storage medium for identifying user intention based on text classification
CN104463101B (en) Answer recognition methods and system for character property examination question
CN111523621A (en) Image recognition method and device, computer equipment and storage medium
CN111444952A (en) Method and device for generating sample identification model, computer equipment and storage medium
CN108460649A (en) A kind of image-recognizing method and device
CN105844283A (en) Method for identifying category of image, image search method and image search device
CN106447625A (en) Facial image series-based attribute identification method and device
CN113869398B (en) Unbalanced text classification method, device, equipment and storage medium
CN111062424A (en) Small sample food image recognition model training method and food image recognition method
CN109345553A (en) A kind of palm and its critical point detection method, apparatus and terminal device
CN109685104B (en) Determination method and device for recognition model
CN108229692B (en) Machine learning identification method based on dual contrast learning
CN109858476A (en) The extending method and electronic equipment of label
CN116051917B (en) Method for training image quantization model, method and device for searching image
CN114066438B (en) Model-based monitoring data display method, device, equipment and storage medium
CN114067155A (en) Image classification method, device, product and storage medium based on meta learning
US11893773B2 (en) Finger vein comparison method, computer equipment, and storage medium
CN112308149A (en) Optimization method and device for image information identification based on machine learning
CN119541552A (en) A complex emotion representation and perception method based on emotion vector
CN105956629A (en) Mode classification method and mode classification system
CN119830149A (en) Radiation source open set identification method and device based on dynamic group constant-change network and generation countermeasure network
CN106355066A (en) Face authentication method and face authentication device
CN108345943B (en) Machine learning identification method based on embedded coding and contrast learning

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20160203

RJ01 Rejection of invention patent application after publication