CN106372552B - Human body target recognition positioning method - Google Patents

Abstract

The present invention provides a method for identifying and positioning a human body target, comprising: obtaining a first positioning result of positioning the human body target by using an ultra-high frequency radio frequency identification system; obtaining a second positioning result of positioning the human body target by using a computer vision system; Suppose a fusion algorithm fuses the first positioning result and the second positioning result to obtain a third positioning result for the human target; using a preset fusion optimization algorithm to optimize the positioning accuracy of the third positioning result to obtain the human target the final positioning result. The invention can realize the identification and positioning of the human body target, and has high efficiency and high accuracy.

Description

Human target recognition and positioning method

technical field

The invention relates to the technical field of security identification and positioning, in particular to a method and device for identifying and positioning a human body target.

Background technique

At present, the demand for security and the security industry arising from it have always been accompanied by the development of human civilization. The application of Closed Circuit Television (CCTV) in New York City Security in the 1960s marked the birth of the modern security industry. Since then, with the development of technology and increasingly severe social security risks, many technologies have been applied to the security industry and a series of security equipment have been derived. Traditionally, a complete security system consists of three parts: a monitoring module, a risk assessment module and a response module. Although the popularity of intelligent technology has blurred the boundaries of the three, there is no doubt that the monitoring module is still the forefront of the entire system. The sensitivity of the monitoring equipment and the ability to gather information determine the accuracy and robustness of the entire security equipment.

However, due to the technical weaknesses of the three commonly used detection equipment: closed-circuit monitoring system (ie computer vision system), motion detector and UHF radio frequency identification system, any single monitoring system cannot meet the requirements of modern security industry for efficiency, accuracy and automation requirements. In recent years, in addition to researching new monitoring methods, various fusion schemes of multi-monitoring systems have been proposed. Most of these solutions focus on the fusion of the final results of various monitoring subsystems, and such higher-level fusion is often difficult to fully utilize the advantages of each subsystem. Therefore, it is necessary to propose a deeper fusion method of multi-monitoring methods to overcome the limitations of the existing schemes and improve the efficiency of the fusion system.

In view of this, how to provide a high-efficiency and high-accuracy human target recognition and positioning method has become a technical problem that needs to be solved at present.

SUMMARY OF THE INVENTION

In order to solve the above technical problems, the present invention provides a human body target identification and positioning method, which can realize the identification and positioning of the human body target with high efficiency and high accuracy.

In a first aspect, the present invention provides a method for identifying and locating a human body target, comprising:

Obtain the first positioning result of positioning the human body target by using the ultra-high frequency radio frequency identification system;

obtaining a second localization result of using the computer vision system to locate the human target;

Using a preset fusion algorithm to fuse the first positioning result and the second positioning result to obtain a third positioning result for the human target;

A preset fusion optimization algorithm is used to optimize the positioning accuracy of the third positioning result to obtain a final positioning result for the human target.

Optionally, the obtaining the first positioning result of locating the human target by using the ultra-high frequency radio frequency identification system includes:

Using the angle-of-arrival positioning algorithm based on the passive label column, the position of the label column on the human body target in the UHF radio frequency identification system is located, and then the first positioning result of locating the human body target is obtained;

Wherein, the UHF radio frequency identification system includes: a label column on the human body target and at least three UHF radio frequency identification devices arranged at the same straight line interval; any adjacent two UHF radio frequency identification devices are pre-spaced. Assuming a first distance, the label column includes: two electronic labels separated by a preset second distance.

Optionally, the preset second distance is greater than zero and less than λ/4, where λ is the wavelength of the carrier wave sent by the UHF radio frequency identification device.

Optionally, the position of the label column on the human body target in the UHF radio frequency identification system is located by using the angle of arrival positioning algorithm based on the passive label column, and then the first positioning result of the positioning of the human body target is obtained, include:

Obtain the angle of arrival of each UHF RFID device receiving the echo signal returned by the tag column through its antenna;

According to the geometric relationship between the angle of arrival, the position of the label column and the antenna position of the UHF RFID device, perform two-dimensional positioning on the position of the label column, and then obtain the first method for locating the human target. A positioning result.

Optionally, the obtaining of the second positioning result of positioning the human target by using a computer vision system includes:

Obtain the images collected after the camera set at the preset position in the computer vision system shoots the human target;

Utilize the Gaussian mixture model algorithm to count the change of the pixel value of each pixel in the image;

Using the directional gradient histogram algorithm to detect the human target in the image;

The coordinate system transformation of the human body target detected in the image is performed to obtain the actual world coordinates of the human body target, and then the second positioning result of the real position of the human body target is obtained.

Optionally, performing coordinate system transformation on the human target detected in the image to obtain the actual world coordinates of the human target, and then obtaining a second positioning result of the real position of the human target, including:

A three-dimensional Cartesian coordinate system is established with the position of the camera as the origin, and the three unit coordinate vectors of the three-dimensional Cartesian coordinate system are used to represent the unit coordinate vector of the world coordinate system. The coordinates are mapped to the transfer matrix of the actual plane coordinates of the human body target, and the actual world coordinates of the human body target are calculated according to the transfer matrix, and then the second positioning result of the real position of the human body target is obtained.

Optionally, the detection of the human target in the image by using a directional gradient histogram algorithm includes:

converting the image to a grayscale image;

Normalize the pixel values in the grayscale image using a Gamma correction algorithm;

Calculate the gradient direction and size of each pixel;

Divide the normalized grayscale image into square units with the same size, and count the gradient direction and size of each pixel, so as to obtain the feature vector of each square unit;

Combine multiple adjacent square units into a rectangular block, and normalize the feature vectors in the rectangular block to obtain the feature descriptor of the rectangular block;

The feature descriptors of all blocks are combined to obtain the gradient histogram feature vector of the image, and then the human target in the image is detected.

Optionally, using a preset fusion algorithm to fuse the first positioning result and the second positioning result to obtain a third positioning result for the human target, including:

Using the variance weighted average algorithm, the first positioning result and the second positioning result are preliminarily fused;

The Kalman filtering algorithm is used to estimate the true value of the signal according to the measurement data, to further improve the accuracy of the positioning result obtained after the preliminary fusion, and obtain the third positioning result of the human target.

Optionally, the preset fusion optimization algorithm includes: an initialization process, a cycle tracking process and a display process;

The initialization process is used to initialize a captured new human target;

The cycle tracking process is used to continuously update and track the position of the human target;

The display process is used to draw the result of the loop tracking in each frame of image;

Wherein, the initialization process and the cycle tracking process both include: a preprocessing module, a data comparison module, and a precision enhancement module; wherein:

Described preprocessing module: according to the position of the human body target obtained by the first positioning result or the position of the previous frame of the human body target to estimate the area where the human body target may exist at present;

The data comparison module: compares the target position of the human body obtained by the first positioning result and the target position of the human body obtained by the second positioning result with the previous target position of the human body, selects the most likely target position of the human body and sends it to Precision enhancement module;

The accuracy enhancement module is configured to optimize the accuracy of the third positioning result according to the most probable human target position selected by the data comparison module to obtain a final positioning result for the human target.

As can be seen from the above technical solutions, the human body target identification and positioning method of the present invention uses a preset fusion algorithm to combine the first positioning result of positioning the human body target by using the ultra-high frequency radio frequency identification system and the first positioning result of using the computer vision system to locate the human body target. Two positioning results are fused, and a preset fusion optimization algorithm is used to optimize the positioning accuracy of the fusion results to obtain the final positioning result of the human target, thereby realizing the identification and positioning of the human target with high efficiency and high accuracy.

Description of drawings

1 is a schematic flowchart of a method for identifying and positioning a human body target provided by an embodiment of the present invention;

2 is a schematic diagram of a basic model of an AOA positioning algorithm based on a passive label column provided by an embodiment of the present invention;

3 is a schematic diagram of the geometric relationship between the angle of arrival, the position of the tag column, and the position of the antenna in the angle of arrival positioning algorithm based on the passive label column provided by an embodiment of the present invention;

FIG. 4 is the antenna placement of the verification and evaluation experiment of the first positioning result obtained by using the passive tag column-based angle of arrival positioning algorithm to locate the label column position on the human body target in the UHF RFID system according to the embodiment of the present invention. schematic diagram;

5a is a schematic flowchart of a specific process of a method for identifying and positioning a human body target provided by an embodiment of the present invention;

FIG. 5b is a schematic flowchart of a specific process of a method for identifying and positioning a human body target provided by an embodiment of the present invention.

Detailed ways

In order to make the purposes, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments It is only a part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

Table 1 shows the comparison of the advantages and disadvantages of the UHF RFID system, the computer vision system and the motion detector. From Table 1, it can be seen that only the computer vision system and the UHF RFID system can locate human targets in a large area. ability. This commonality means that the two systems can achieve deep fusion by sharing the location information of human objects.

Table 1

FIG. 1 shows a schematic flowchart of a method for identifying and positioning a human body target provided by an embodiment of the present invention. As shown in FIG. 1 , the method for identifying and positioning a human body target in this embodiment includes the following steps 101-104.

101. Acquire a first positioning result of positioning a human target by using an ultra-high frequency radio frequency identification system.

102. Acquire a second positioning result of positioning the human target by using the computer vision system.

103. Use a preset fusion algorithm to fuse the first positioning result and the second positioning result to obtain a third positioning result for the human target.

104. Use a preset fusion optimization algorithm to optimize the positioning accuracy of the third positioning result to obtain a final positioning result for the human body target.

In the human body target identification and positioning method of this embodiment, a first positioning result obtained by using an ultra-high frequency radio frequency identification system to locate a human body target and a second positioning result obtained by using a computer vision system to locate the human body target are fused by a preset fusion algorithm. , and use the preset fusion optimization algorithm to optimize the positioning accuracy of the fusion result, and obtain the final positioning result of the human target, thereby realizing the identification and positioning of the human target with high efficiency and high accuracy.

In a specific application, the above step 101 may include:

Using the angle-of-arrival positioning algorithm based on the passive label column, the position of the label column on the human body target in the UHF radio frequency identification system is located, and then the first positioning result of locating the human body target is obtained;

Wherein, the UHF radio frequency identification system includes: a label column on the human body target and at least three UHF radio frequency identification devices arranged at the same straight line interval; any adjacent two UHF radio frequency identification devices are pre-spaced. Assuming a first distance, the label column includes: two electronic labels separated by a preset second distance.

Wherein, the preset second distance is greater than zero and less than λ/4, where λ is the wavelength of the carrier wave sent by the UHF RFID device.

In a specific application, the UHF radio frequency identification device is an UHF radio frequency identification RFID reader.

Specifically, by using the angle-of-arrival positioning algorithm based on the passive tag column to locate the position of the label column on the human body target in the UHF radio frequency identification system, and then to obtain the first positioning result of locating the human body target, it is possible to It further specifically includes steps 101a and 101b not shown in the figure:

101a. Acquire the angle of arrival at which each UHF RFID device receives the echo signal returned by the tag column through its antenna.

Specifically, the step 101a may include steps S1 and S2 not shown in the figure:

S1. Acquire the arrival phase offset values of the two echo signals respectively returned by the two electronic tags in the tag column received by each UHF radio frequency identification device through its antenna.

S2. According to the arrival phase offset value of the two echo signals, the wavelength of the carrier wave sent by the UHF RFID device, and the preset second distance, obtain each UHF RFID device through its The antenna receives the angle of arrival of the echo signal returned by the tag column.

Further, the step S2 may specifically include:

According to the arrival phase offset value of the two echo signals and The wavelength λ of the carrier wave sent by the UHF RFID device and the preset second distance d are obtained through the first formula to obtain the echoes returned by each UHF RFID device through its antenna to receive the tag column. The arrival angle θ of the signal;

Wherein, the first formula is:

101b. According to the geometric relationship between the angle of arrival, the position of the label column, and the position of the antenna of the UHF RFID device, perform two-dimensional positioning on the position of the label column, and then obtain the positioning of the human body target the first positioning result.

Specifically, the derivation process of the above formula (1) is as follows:

Referring to FIG. 2, FIG. 2 shows the basic model of the angle-of-arrival positioning algorithm based on the passive tag column, the basic model includes: the antenna C of any UHF RFID device in the UHF RFID system and the For the label column (that is, two electronic labels A and B separated by a preset second distance), let the coordinates of the electronic label A be (0, 0), the coordinates of the electronic label B be (d, 0), and the coordinates of the antenna C The coordinates are (x, h), then the difference between the distances between the electronic tags A and B and the antenna C is:

Since d is much smaller than the distance from the antenna C to the electronic tag A/B, the infinitesimal substitution on the right side of the formula (2) can be obtained:

Considering the linear relationship between the arrival phase offset and the distance of each UHF RFID device receiving the echo signal returned by the tag column through its antenna, we can get:

in, is the arrival phase offset value of the two echo signals respectively returned by the electronic tags A and B received by the UHF RFID device through the antenna C;

make According to formulas (3) and (4), the transformation can be obtained:

Since the label column introduces a new degree of freedom (the direction of the label column), the above step 101b requires three or more antennas to realize the two-dimensional positioning of the label column position. The two-dimensional coordinates of the tag column can be calculated according to the obtained angle of arrival according to a simple geometric relationship. FIG. 3 is a schematic diagram of the geometric relationship among the angle of arrival, the position of the tag column and the position of the antenna.

It should be noted that the selection process of the preset second distance d is as follows:

On the one hand, in formula (5) The range of values is [-1,1], and The value range of is [0, 2π), so formula (5) is valid for any Both hold if and only if d<λ/4. In a specific application, for UHF RFID equipment, the wavelength of the carrier wave is about 32cm, so d should be less than 8cm. On the other hand, the following error analysis shows that the random error of positioning is inversely proportional to the electronic tag interval d, so d should be as large as possible to reduce the random error. Therefore, the value of d in this embodiment should preferably be 8 cm.

In order to verify and evaluate the angle-of-arrival algorithm based on the passive tag column, this example uses an UHF RFID reader with the signal IPJ-REV-R420-GX21M for testing, the antenna model is E911011PCR, and its directional gain is 11dBic , the beam width is 40°, and the antenna placement method is shown in Figure 4. A1 and A2 are the antennas, and b and b' are the circumferences of the antennas. The label columns are placed within the rectangle D1D2D3D4, and the line spacing between the label columns is 0.5m. The experimental results show that the maximum positioning deviation is 0.43m, and the minimum positioning deviation is 0.07m. This result meets the theoretical expectation, and the detection result of the method described in this embodiment has high accuracy.

In a specific application, the above step 102 may include steps 102a-102d not shown in the figure:

102a: Acquire an image collected after a camera set at a preset position in a computer vision system shoots a human target.

102b. Use a Gaussian mixture model algorithm to count changes in the pixel value of each pixel in the image.

It should be noted that, in the Gaussian mixture model algorithm, K Gaussian distribution models need to be established for each pixel, and these Gaussian distribution models are used to count the changes in the pixel value of each pixel in the image, that is, the pixel value in the image. Each pixel uses K Gaussian distribution models to represent its value probability in the time domain;

The probability that the current pixel value is x _t =[x _1,t ,x _2,t ,...,x _n,t ] can be expressed as:

Among them, K is the number of Gaussian models, w _k,t is the weight of the kth Gaussian distribution model in the current pixel at time t, μ _k,t is the mathematical expectation of the kth Gaussian distribution model in the current pixel at time t, σ _{k, t} is the covariance matrix of the kth Gaussian distribution model in the current pixel at time t, and n is a positive integer. For simplicity, it is usually considered that the three co-channels in the RGB color space or the YUV color space are independent of each other, namely where I is a three-dimensional unit matrix, σ _{k, t} is the standard deviation of each color component in the kth Gaussian distribution model in the current pixel at time t.

The Gaussian distribution model of each pixel is sorted according to w _k,t /σ _k,t , and this value can represent the probability that the corresponding Gaussian distribution model is the background model. If a pixel value does not satisfy any of the Gaussian distribution models, a new Gaussian distribution model will be established, and the lowest ranked Gaussian distribution model will be eliminated.

In order to further simplify the operation, at time t, the pixel value x _t is assigned to the Kth Gaussian distribution model if and only if it satisfies the following formula (3):

|x _t -μ _k,t-1 |＜Dσ _k,t-1 (8)

Among them, D is a user-defined parameter, usually the value is 2.5.

At this time, the parameters of the Gaussian Gaussian distribution model are updated recursively as follows:

w _k,t =(1-ρ)w _k,t-1 +ρ (9)

μ _k,t =(1-ρ)μ _k,t-1 +ρx _t (10)

Among them, ρ is the learning rate, which can be between 0 and 1. The value of ρ determines the rate at which the Gaussian distribution model is updated.

The weights of the unmatched Gaussian distribution model are updated according to the following formula (12):

w _i,t =(1-α)wi _,t-1 i≠k (12)

If the pixel value does not satisfy any Gaussian distribution model, a new Gaussian distribution model will be established. The expectation of the new Gaussian distribution model is the pixel value x _t , and its standard deviation and expectation will be set to preset default values.

102c, using a directional gradient histogram algorithm to detect a human target in the image.

Specifically, the step 102c may include steps A1-A6 not shown in the figure:

A1. Convert the image into a grayscale image.

A2. Use a gamma correction algorithm to normalize the pixel values in the grayscale image.

In specific applications, the expression of the Gamma correction algorithm is:

I'(x,y)=I(x,y) ^Γ (13)

Among them, I(x, y) is the input pixel value, I'(x, y) is the output pixel value, Γ is the gamma parameter, and the value range of Γ is (0, 1), usually Γ takes The value is 0.5.

It can be understood that step A2 can reduce the influence from illumination variation and random noise.

A3. Calculate the gradient direction and size of each pixel.

In specific applications, the formulas for calculating the gradients G _x and G _y of each pixel are:

Among them, G _x (x, y) is the gradient vector in the horizontal direction of the pixel with coordinates (x, y), and G _y (x, y) is the gradient of the pixel with coordinates (x, y) in the horizontal direction. vector, H(x, y) is the pixel value of the pixel whose coordinates are (x, y);

The modulo G(x,y) of the gradient (ie the gradient magnitude) is:

The gradient direction θ'(x,y) is:

Typically, this step can be achieved by two-dimensional convolution of the image matrix with the Canny edge detection operator.

A4. Divide the normalized grayscale image into square units with the same size, and count the gradient direction and size of each pixel, so as to obtain the feature vector of each square unit.

A5. Combine a plurality of adjacent square units into a rectangular block, and normalize the feature vectors in the rectangular block to obtain a feature descriptor of the rectangular block.

A6. Combine the feature descriptors of all blocks together to obtain a gradient histogram feature vector of the image, and then detect a human target in the image.

102d. Perform coordinate system transformation on the human body target detected in the image to obtain the actual world coordinates of the human body target, and then obtain a second positioning result of the real position of the human body target.

Specifically, the step 102d may include:

A three-dimensional Cartesian coordinate system is established with the position of the camera as the origin, and the three unit coordinate vectors of the three-dimensional Cartesian coordinate system are used to represent the unit coordinate vector of the world coordinate system. The coordinates are mapped to the transfer matrix of the actual plane coordinates of the human body target, and the actual world coordinates of the human body target are calculated according to the transfer matrix, and then the second positioning result of the real position of the human body target is obtained.

It should be noted that, in this embodiment, the angle between the pointing of the camera and the horizontal plane can be regarded as the angle between the pointing of the camera and a straight line determined by the position of the camera and a point along the vector V to infinity. The distance L from the image point to the center of the image (also called the vanishing point) is proportional to the tangent of the angle between the line connecting the vanishing point and the imaged point and the camera pointing [4]. Its expression is given by the following equations (17) and (18):

Among them, α is the horizontal angle formed by the imaging center and the imaged point, β is the pitch angle formed by the imaging center and the imaged point, and x' and z' are the horizontal and vertical pixel distances from the image point to the center of the imaging plane.

As mentioned above, the key to estimating the camera's pointing is to find the pixel distance between the center of the image and a "point" at infinity on the axes of the world coordinate system. And the position of the "point" at infinity in the image can be obtained by calculating the "intersection point" of two straight lines that are actually parallel in the vertical or horizontal plane of the world coordinate system.

Usually in indoor monitoring, the joint between the floor or ceiling can be regarded as an ideal reference parallel line, so the position on the image of the "point" at infinity can be obtained through the following steps:

a) Use Hough transform to find all the straight lines in the image;

b) Select a straight line whose length meets the requirements;

c) Filter out all straight lines whose slopes do not meet the requirements according to the prior attitude information from the gimbal;

d) Merge two straight lines that are too close into one

e) Find the point where three or more straight lines pass in common.

A three-dimensional Cartesian coordinate system with the camera position as the origin is established, which is called the camera coordinate system. Let the camera point to be the coordinate axis V, let the horizontal right axis be the coordinate axis U, and the coordinate axis W is given by U×V. The three axes of the three-dimensional world coordinate system are X, Y, and Z. According to the camera pointing information obtained in the previous section, the unit coordinate vector of the world coordinate system can be replaced by the three unit vectors of the camera coordinate system. Expressed as:

So there are:

On this basis, the transition matrix that maps the target pixel coordinates to the actual plane coordinates of the target is further obtained.

Assume that the Z coordinate of the human target is a fixed value h. Since the mapping from camera coordinates to world coordinates is a linear mapping, the horizontal plane z=h in the world coordinate system must be mapped as a plane in the camera coordinate system. Suppose this plane is

v=au+bw+c (22)

According to formula (20), we can get

then

Obviously the right-hand part of the equal sign of Equation (24) should be independent of u or w, so the equation

Solve to get:

Therefore, the actual coordinate [x, y] ^T can be represented by the coordinate [u, w] ^T on the graph as

Likewise, the coordinates on the image can be represented by the actual coordinates as

Thereby, the actual coordinates of the specific person identification are obtained through the matrix calculation.

In a specific application, the above step 103 may include steps 103a and 103b not shown in the figure:

103a. Perform preliminary fusion of the first positioning result and the second positioning result by using a variance weighted average algorithm.

In a specific application, the process of the variance weighted average algorithm may include:

It is assumed that the position measurement equations of the UHF RFID system and the computer vision system are:

B=X+U (31)

C=X+V (32)

Among them, B is the position vector measured by the UHF RFID system; C is the position vector measured by the computer vision subsystem; X is the actual position vector; U and V are noise vectors, and the two satisfy:

where Q and R are both positive definite matrices;

Let Z=K×B+(I-K)×C (35)

Among them, K is the coefficient matrix when the Z covariance matrix is the smallest, which is called the optimal coefficient matrix;

The offset between Z and the actual position X is:

The covariance matrix of Z is:

make If the optimal coefficient matrix K is added to the offset coefficient matrix δK, we have

δP=δKW+(δKW) ^T +δK(Q+R)δK ^T (39)

in:

W=QK ^T -R(IK) ^T (40)

According to the previous assumption, (Q+R) is a positive definite matrix. Therefore, no matter what the value of δK is, the third term of formula (39) is always positive; and when W≠0, the sign of the first two terms of formula (39) changes with the change of δK. Therefore, the variance of P is the smallest if and only if W≡0, i.e.

K=R(Q+R) ^-1 (41)

(IK)=Q(Q+R) ^-1 (42)

At this time, the matrix of Z and its covariance is:

Z=R(Q+R) ^-1 ×B+Q(Q+R) ^-1 ×C (43)

P=(Q+R) ^-1 (QRQ+RQR)(Q+R) ^-1 (44)

The determinant of the covariance matrix is:

|P| is less than either of |Q| and |R|.

103b. Use the Kalman filter algorithm to estimate the true value of the signal according to the measurement data, further improve the accuracy of the positioning result obtained after the preliminary fusion, and obtain a third positioning result for the human target.

In a specific application, the process of the Kalman filter algorithm may include:

The state vector at time t _k is defined as X _k . X _k is driven by random noise sequence W _k-1 , and its driving equation is:

X _k =Φ _k,k-1 X _k-1 +Γ _k-1 W _k-1 (46)

The measurement equation is:

Z _k =H _k X _k +V _k (47)

Among them, Φ _k,k-1 is the transition matrix from time t _k-1 to time t _k ; Γ _k-1 is the driving matrix; H _k is the measurement matrix; V _k is the random measurement noise.

The random noise sequences W _k and V _k satisfy the following relationship:

Estimated value of the state vector X _k It can be obtained according to the following recursion:

One-step state estimation:

State estimation:

Filter gain:

One-step state estimation variance:

State estimate variance:

In a specific application, the preset fusion optimization algorithm in the above step 104 may include: an initialization process, a loop tracking process, and a display process;

The initialization process is used to initialize a captured new human target;

The cycle tracking process is used to continuously update and track the position of the human target;

The display process is used to draw the result of the loop tracking in each frame of image;

Wherein, the initialization process and the cycle tracking process both include: a preprocessing module, a data comparison module, and a precision enhancement module; wherein:

Described preprocessing module: according to the position of the human body target obtained by the first positioning result or the position of the previous frame of the human body target to estimate the area where the human body target may exist at present;

The data comparison module: compares the target position of the human body obtained by the first positioning result and the target position of the human body obtained by the second positioning result with the previous target position of the human body, selects the most likely target position of the human body and sends it to Precision enhancement module;

The accuracy enhancement module is configured to optimize the accuracy of the third positioning result according to the most probable human target position selected by the data comparison module to obtain a final positioning result for the human target.

In order to evaluate the effectiveness of the preset fusion optimization algorithm, this embodiment uses a surveillance video recorded by a commercial-grade closed-circuit surveillance camera (DS-2DC2202-DE3/W HIKVISION) to conduct experiments. Due to the complexity of the electromagnetic environment of the experimental site, it is difficult to meet the needs of UHF RFID positioning. The simulation results are used for the positioning data of UHF RFID here. The length of the video is about 6 seconds, with a total of 130 frames. The results show that the false detection phenomenon is completely eliminated, while the positioning accuracy of the target is improved by about an order of magnitude: the root mean square error is improved from 0.0673m for the UHF RFID system and 0.1226m for the computer vision system to 0.0107m. The output results of Kalman filtering show that the absolute error has been further improved compared with the initial fusion results, the root mean square value is 0.0071m, the improvement range is about 30%, and the position of the target is accurately marked.

FIG. 5a and FIG. 5b respectively show a specific schematic flowchart of a method for identifying and positioning a human body target provided by an embodiment of the present invention.

In the human body target identification and positioning method of this embodiment, a first positioning result obtained by using an ultra-high frequency radio frequency identification system to locate a human body target and a second positioning result obtained by using a computer vision system to locate the human body target are fused by a preset fusion algorithm. , and use the preset fusion optimization algorithm to optimize the positioning accuracy of the fusion result, and obtain the final positioning result of the human target, which can realize the identification and positioning of the human target, with high efficiency and high accuracy.

As will be appreciated by those skilled in the art, the embodiments of the present application may be provided as a method, a system, or a computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It will be understood that each flow and/or block in the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing device produce Means for implementing the functions specified in a flow or flow of a flowchart and/or a block or blocks of a block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions The apparatus implements the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.

These computer program instructions can also be loaded on a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process such that The instructions provide steps for implementing the functions specified in the flow or blocks of the flowcharts and/or the block or blocks of the block diagrams.

It should be noted that, in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any relationship between these entities or operations. any such actual relationship or sequence exists. Moreover, the terms "comprising", "comprising" or any other variation thereof are intended to encompass a non-exclusive inclusion such that a process, method, article or device that includes a list of elements includes not only those elements, but also includes not explicitly listed or other elements inherent to such a process, method, article or apparatus. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in a process, method, article or apparatus that includes the element. The orientation or positional relationship indicated by the terms "upper", "lower", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the indicated device or element must be It has a specific orientation, is constructed and operates in a specific orientation, and therefore should not be construed as a limitation of the present invention. Unless otherwise expressly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection, It can also be an electrical connection; it can be a direct connection, an indirect connection through an intermediate medium, or an internal connection between two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

In the description of the present invention, numerous specific details are set forth. It will be understood, however, that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description. Similarly, it is to be understood that, in the above description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together into a single embodiment in order to simplify the present disclosure and to aid in the understanding of one or more of the various aspects of the invention. , figures, or descriptions thereof. However, this method of disclosure should not be construed to reflect the intention that the invention as claimed requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the Detailed Description are hereby expressly incorporated into this Detailed Description, with each claim standing on its own as a separate embodiment of this invention. It should be noted that the embodiments in the present application and the features of the embodiments may be combined with each other in the case of no conflict. The invention is not limited to any single aspect, nor to any single embodiment, nor to any combination and/or permutation of these aspects and/or embodiments. Furthermore, each aspect and/or embodiment of the invention may be used alone or in combination with one or more other aspects and/or embodiments thereof.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, but not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: The technical solutions described in the foregoing embodiments can still be modified, or some or all of the technical features thereof can be equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the embodiments of the present invention. The scope of the invention should be included in the scope of the claims and description of the present invention.

Claims (8)

1. A human body target identification and positioning method is characterized by comprising the following steps:

acquiring a first positioning result of positioning a human body target by using an ultrahigh frequency radio frequency identification system;

acquiring a second positioning result for positioning the human body target by using a computer vision system;

fusing the first positioning result and the second positioning result by using a preset fusion algorithm to obtain a third positioning result of the human body target;

optimizing the positioning precision of the third positioning result by using a preset fusion optimization algorithm to obtain a final positioning result of the human body target;

wherein, the preset fusion optimization algorithm comprises: an initialization process, a loop tracking process and a display process;

the initialization process is used for initializing a captured new human body target;

the cyclic tracking process is used for continuously updating and tracking the position of the human body target;

the display process is used for drawing the result of the circular tracking in each frame of image;

wherein the initialization process and the loop tracking process both comprise: the device comprises a preprocessing module, a data comparison module and a precision enhancement module; wherein:

the preprocessing module is used for: estimating the possible existing area of the current human body target according to the position of the human body target obtained by the first positioning result or the position of the last frame of the human body target;

the data comparison module: comparing the human body target position obtained by the first positioning result and the human body target position obtained by the second positioning result with the previous human body target position respectively, selecting the most possible human body target position and sending the most possible human body target position to the precision enhancing module;

and the precision enhancing module is used for carrying out precision optimization on the third positioning result according to the most possible human body target position selected by the data comparison module to obtain a final positioning result of the human body target.

2. The method of claim 1, wherein obtaining a first positioning result of positioning the human target using an UHF RFID system comprises:

positioning the position of a tag column on a human body target in an ultrahigh frequency radio frequency identification system by using an arrival angle positioning algorithm based on a passive tag column so as to obtain a first positioning result for positioning the human body target;

wherein, the hyperfrequency radio frequency identification system includes: the system comprises a label column on a human body target and at least three ultrahigh frequency radio frequency identification devices which are arranged on the same straight line at intervals; any two adjacent ultrahigh frequency radio frequency identification devices are separated by a preset first distance, and the label column comprises: two electronic tags separated by a preset second distance.

3. The method of claim 2, wherein the predetermined second distance is greater than zero and less than λ/4, λ being a wavelength of a carrier wave transmitted by the UHF RFID device.

4. The method of claim 3, wherein the positioning the position of the tag column on the human body target in the UHF RFID system by using the passive tag column-based angle-of-arrival positioning algorithm to obtain the first positioning result for positioning the human body target comprises:

obtaining an arrival angle of each ultrahigh frequency radio frequency identification device for receiving an echo signal returned by the label column through an antenna of the ultrahigh frequency radio frequency identification device;

and according to the geometric relation among the arrival angle, the positions of the label columns and the antenna position of the ultrahigh-frequency radio frequency identification device, performing two-dimensional positioning on the positions of the label columns, and further obtaining a first positioning result for positioning the human body target.

5. The method of claim 1, wherein obtaining a second positioning result for positioning the human target using the computer vision system comprises:

acquiring an image acquired after a camera arranged at a preset position in a computer vision system shoots a human body target;

counting the change of the pixel value of each pixel in the image by using a Gaussian mixture model algorithm;

detecting a human body target in the image by using a direction gradient histogram algorithm;

and converting a coordinate system of the human body target detected in the image to obtain the actual world coordinate of the human body target, and further obtain a second positioning result of the actual position of the human body target.

6. The method according to claim 5, wherein the converting the coordinate system of the human target detected in the image to obtain the actual world coordinates of the human target and further obtain the second positioning result of the real position of the human target comprises:

establishing a three-dimensional Cartesian coordinate system by taking the position of the camera as an origin, expressing a unit coordinate vector of a world coordinate system by using three unit coordinate vectors of the three-dimensional Cartesian coordinate system, further obtaining a transfer matrix for mapping the pixel coordinates of the human body target to the actual plane coordinates of the human body target on the basis, calculating the actual world coordinates of the human body target according to the transfer matrix, and further obtaining a second positioning result of the actual position of the human body target.

7. The method of claim 5, wherein the detecting the human target in the image by using a direction gradient histogram algorithm comprises:

converting the image into a grey-scale map;

normalizing the pixel values in the gray-scale image by using a Gamma correction algorithm;

calculating the gradient direction and the size of each pixel point;

dividing the normalized gray level image into square units with the same size, and counting the gradient direction and size of each pixel to obtain a feature vector of each square unit;

combining a plurality of adjacent square units into a rectangular block, and normalizing the feature vectors in the rectangular block to obtain a feature descriptor of the rectangular block;

combining the feature descriptions of all the blocks together to obtain a gradient histogram feature vector of the image, and further detecting a human body target in the image.

8. The method according to claim 1, wherein the fusing the first positioning result and the second positioning result by using a preset fusion algorithm to obtain a third positioning result of the human body target comprises:

performing preliminary fusion on the first positioning result and the second positioning result by using a variance weighted average algorithm;

and further improving the precision of the positioning result obtained after the preliminary fusion by using a Kalman filtering algorithm according to the true value of the measurement data estimation signal, and obtaining a third positioning result of the human body target.