CN109002821B - A digital identification method of online banking shield based on connected domain and tangent slope - Google Patents

Abstract

The invention discloses a digital identification method for online banking shields based on connected domains and tangent slopes. The expansion structure element is used to eliminate laser digital string breakpoints. In order to ensure that the breakpoints are completely eliminated, a larger expansion factor is set, so that the digital string has some Adhesion, the extremum segmentation method is performed on the glued number string to separate the numbers, and finally, 6 geometric and topological features are used to construct a digital prototype to distinguish the numbers, so as to identify the numbers. The actual digital recognition experiment of online banking shield shows that the recognition accuracy rate of this recognition method for digital strings is as high as 99%, and the average recognition time is 0.15s. The system has high recognition accuracy and has a certain real-time performance. Further improvement of the system can be applied to online monitoring of laser printing production lines, as well as consistency verification of products and packaging before products leave the factory.

Description

A digital identification method of online banking shield based on connected domain and tangent slope

technical field

The invention belongs to the technical field of automatic identification and detection, and relates to a digital identification method for an online banking shield, in particular to a digital identification method for an online banking shield based on a connected domain and a tangent slope.

Background technique

Compared with the traditional printed digital strings with higher quality of printed products, the laser printed digital strings cannot be accurately segmented and identified by traditional algorithms due to the problems of breakpoints and rough edges.

SUMMARY OF THE INVENTION

In order to solve the above technical problems, the present invention provides a digital identification method for online banking shields based on connected domains and tangent slopes.

The technical scheme adopted in the present invention is: a digital identification method based on the connected domain and the tangent slope, which is characterized in that it comprises the following steps:

Step 1: Preprocess the image;

Step 2: Divide the digital string;

Including broken character flattening and glued character extraction;

Step 3: Number string identification;

Including recognition feature selection and character classification recognition.

Relative to the prior art, the beneficial effects of the present invention are:

(1) An algorithm for segmentation and recognition of laser printed digital strings is proposed. Compared with the existing digital recognition methods, this algorithm proposes a new expansion to eliminate breakpoints for laser numbers with poor printing quality. In the method of segmentation by extreme value method, six geometric and topological features are used to construct digital prototypes in the database, and characters are distinguished and recognized by comparing the segmentation results of digital strings with the digital prototypes.

(2) The experimental results show that the algorithm has high recognition accuracy, high speed, and certain real-time performance. It has a good ability to recognize laser fractured and glued characters, and can recognize slightly inclined characters. The algorithm's recognition accuracy for 12-digit strings is as high as 99%, and the average recognition time is 0.15s.

Description of drawings

1 is a flowchart of an embodiment of the present invention;

2 is a schematic diagram of an image binarization result according to an embodiment of the present invention;

3 is a horizontal and vertical projection view of an image according to an embodiment of the present invention;

4 is a comparison diagram of an experiment of digital string expansion processing according to an embodiment of the present invention;

FIG. 5 is a feature diagram of the upper and lower contours of an image according to an embodiment of the present invention;

FIG. 6 is a diagram of a Code 128C code encoding format according to an embodiment of the present invention.

Detailed ways

In order to facilitate the understanding and implementation of the present invention by those of ordinary skill in the art, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the embodiments described herein are only used to illustrate and explain the present invention, but not to limit it. this invention.

Compared with the traditional printed digital strings with higher quality of printed products, the laser printed digital strings cannot be accurately segmented and identified by traditional algorithms due to the problems of breakpoints and rough edges. To this end, the present invention proposes a method for segmenting and identifying a laser-printed digital string. In this method, the expansion structure element is used to eliminate the breakpoint of the laser digital string. In order to ensure the complete elimination of the breakpoint, a larger expansion factor is set, so that the digital string is glued. The digital prototype is constructed with 6 geometric and topological features for distinguishing the numbers, so as to identify the numbers.

Since the laser digital serial number of the online banking shield matches the barcode serial number on the packaging box, the barcode is identified. If the identification results of the two are consistent, the digital identification is successful and the online banking shield is packaged correctly; otherwise, the online banking shield is incorrectly packaged or the serial number If the identification is wrong, the online banking shield needs to be returned to the factory for processing.

See Fig. 1, a kind of online banking shield digital identification method based on connected domain and tangent slope provided by the present invention comprises the following steps:

Step 1: Preprocess the image;

Step 1.1: Image binarization;

Due to the large number of products, the image will be affected by factors such as ambient light and acquisition device parameters, and the gray value of the image will appear different gray value sets. Therefore, in this embodiment, the dynamic threshold processing method Otsu's method is used to perform binarization processing on the image.

The image processing result is shown in Figure 2.

Step 1.2: digital string extraction;

Statistical method is used for the above binarized area, 0 represents the background, 1 represents the boundary of the primitive, and the number of pixels with a gray level of 1 in the horizontal direction is counted, and the pixels whose pixels are lower than the average are removed and the pixels are extracted. After the area with the largest number of points, perform vertical pixel statistics on this area, remove the column with the pixel points lower than the average and extract the area with the most pixel points (the same area with an apparent interval of less than 10 columns) is the area where the target number string is located. The statistical results in the horizontal and vertical directions are shown in Figure 3. The final area height ah is recorded as the character height value. In order to avoid the loss of boundary information, the area boundary is appropriately enlarged.

Step 2: Divide the digital string;

Including broken character flattening and glued character extraction;

Among them, the broken characters are combined, and the specific implementation includes the following sub-steps:

Step 2.1.1: digital string filtering and denoising;

The sampling and transmission of digital images are often disturbed by noise when passing through the sensor or transmission channel, and the dilation algorithm will amplify the noise. Therefore, median filtering is used to eliminate isolated noise points before dilation to keep the edges well.

In this embodiment, median filtering is used to eliminate isolated noise points, and the size of the filtering window is 3×3. The definition of median filter is as follows:

G(m,n)=MedA{fA(m+k,n+l),(k,l)∈w};

Among them, fA(m,n) represents the original image, G(m,n) represents the processed image, m represents the number of rows, and n represents the number of columns; k∈[-1,1], l∈[-1,1] ;w represents a two-dimensional template, which is 3*3;

Step 2.1.2: Remove the breakpoint of the digital string;

In order to eliminate the breakpoints caused by the printing quality of the numbers, this embodiment uses the expansion of the number string; expansion is the operation of merging the background points in contact with the primitives into the primitives to make the primitives in the image "thick". A is inflated by B, defined as:

为结构元，即A被B膨胀是由所有结构元的原点位置组成的集合；In the formula: φ is the empty set,

is a structural element, that is, A is expanded by B, which is a set composed of the origin positions of all structural elements;

This embodiment adopts the expansion structure element as follows:

Dilation processing can eliminate the breakpoints in the digital string very well, and complete the pre-processing of the digital string segmentation. Figure 4 shows the before and after processing of the digital string.

Among them, the glue character extraction is to segment the glue number string, extract a single character, and then normalize the character. The specific implementation includes the following sub-steps:

Step 2.2.1: character segmentation;

Due to the adhesion of some numbers in the number string, the contour extreme value analysis method is used to obtain a reasonable character segmentation point, so as to segment the number string.

The upper and lower contour features are obtained by tracking search, as shown in Figure 5.

The contour function on the image is defined as f(x), where X=(x ₁ , x ₂ ,...,x _n ) ^T is the domain of definition; n is the right boundary of the domain of definition, that is, the total number of points; for X*∈R, If there is a certain ε>0, make all X _a ∈X whose distance from X* is less than ε satisfy the inequality f(X _a )≥f(X*), that is, X _a ∈ X and |X _a -X*| <ε, then X* is called the local minimum point of f(x) on R, and the local minimum value is recorded as f _T (X ^* ); thus the local minimum vector of the contour is X ^T = (X ^T1 , X ^T2 , ..., X ^Tm );

Similarly, the local maximum vector of the lower contour is obtained as X ^B = (X ^B1 , X ^B2 , ..., X ^Bm ); the local maximum value is denoted as f _B (X ^* );

且X_Bj∈[X_Ti-5,X_Ti+5]，且For X _Ti ∈ X _T , if

and X _Bj ∈ [X _Ti -5,X _Ti +5], and

为数字串分割点；记录最终区域高度ah为字符高度值；but

It is the dividing point of the number string; record the final area height ah is the character height value;

After traversing X _T , if D _y =[D ₁ ,D ₂ ,...,D ₁₃ ], the number string segmentation succeeds, otherwise the number string segmentation fails;

Step 2.2.2: character segmentation;

The size of the divided characters is different, and some numbers are slightly inclined due to the processing error of the outer packaging. Therefore, the nearest neighbor interpolation method is used to normalize the image, and the image is processed into a standard image of uniform size to ensure that all characters are The maximum width of the characters is W, and the maximum height is unified to H, so as to reduce the character recognition error caused by different character sizes and improve the recognition accuracy.

Step 2.2.3: Contour mutation point extraction;

Use the contour slope mutation detection algorithm to identify the numbers;

First, the slope between two adjacent points of the projected contour is calculated to determine whether the slope has a sudden change, and the number of slope sudden changes is recorded as a character recognition feature, and the first and last ends of the sudden change are removed during calculation;

The slope between two adjacent points of the bottom contour is:

B(i)=(Bottom(i+1)-Bottom(i))/1;

Among them, i=4,5,...,W-4;

When B(i)>0.15*H, it is recorded as a mutation in the slope, and the number of mutation is recorded as bp;

The slope between two adjacent points on the right contour is:

R(i)=(Right(i+1)-Right(i))/1;

Among them, i=4 _, 5,...,H-4;

In order to improve the recognition accuracy, the left contour is divided into upper and lower parts;

Calculate the slope between two adjacent points on the upper left contour of the image as:

Lu(i)=(Left(i+1)-Left(i))/1;

Among them, i=4,5...,H/2+1;

Calculate the slope between two adjacent points in the lower left contour of the image as:

Ld(i)=(Left(i+1)-Left(i))/1;

Among them, i=H/2+1,...,H-4;

Set the slope threshold to 0.3*W, when any value of R(i), Lu(i), Ld(i) is greater than 0.3*W, it is recorded as a sudden change in the slope, and R(i), Lu(i) are recorded respectively. , The number of contour mutation points in Ld(i) is rp, lup, ldp.

Step 3: Number string identification;

Including recognition feature selection and character classification recognition.

Among them, for the selection of identification features, the extraction of numerical features is used to distinguish characters; in order to accurately classify and identify numbers, six geometric and topological features are selected to construct digital prototypes in the database to distinguish characters. The selected features are as follows:

(1) The number of connected domains cd={1, 2, 3};

This embodiment adopts the 4-connection search method; the 4-connection search method refers to the method of starting from any point on the area and reaching any pixel in the area through the combination of movements in the four directions of up, down, left and right. , the pixels in the left and right directions are respectively denoted as Ii(x, y+1), Ii(x, y-1), Ii(x-1, y), and Ii(x+1, y).

The specific algorithm of the connected domain is as follows:

1) The foreground pixel of the binary image is 1, and the background pixel is 0. Note that the maximum number of connected domains in the current image is Lmax=1, scan the image from top to bottom, and do not process the pixel point Ii(x, y) whose pixel value is 0;

2) Search for the point with the first pixel value of 1, examine Ii(x, y+1), Ii(x, y-1), Ii(x-1, y), Ii(x+1, y ) whether there is at least one value of Lmax, if so, Ii(x,y)=Lmax, otherwise Ii(x,y)=Lmax+1;

3) Take Ii(x, y+1), Ii(x, y-1), Ii(x-1, y), and Ii(x+1, y) as the current point respectively and perform the search in 2) ;

4) Repeat 2) and 3) until the traversal of all points is completed.

After traversing all the points on the character Ii, all connected parts in the character Ii and the number of connected domains Lmax of the character Ii can be obtained, so that the numbers are initially classified according to the number of connected domains: 8 including 3 connected domains, including 2 0, 4, 6, and 9 of the connected domain include 1, 2, 3, 5, and 7 of a connected domain.

(2) The ratio of the maximum width within 3/4 character height to the total character height wd=max{Right(i)-Left(i)}/H;

(3) The number of bottom contour mutation points bp={0, ≥1};

(4) The number of right contour mutation points rp={0, ≥1};

(5) The number of mutation points in the upper left profile lup={0, ≥1};

(6) The number of mutation points in the lower left profile ldp={0, ≥1}.

Suppose the digital prototype S _j , so the number has 10 j∈[0,9], k is the number of features, there are 6 features in total, but S _j with the most number of features is 0, 4 features are selected, and the number of features is obtained The minimum S _j is 8, and 1 feature is selected, so k∈[1,4];

Wherein, the character classification and recognition is to perform character recognition by matching the characters _si obtained after scanning and _segmentation with the digital prototype Si in the database;

where S _j =(f _1j , f _2j ,...,f _mj ), j={1,2,...,k}, m≤6, f _mj is one of the six features in step 3; s _j = ({cd _j },{wd _j },{bp _j },{rp _j },{lup _j },{ldp _j });

且p＝{1,2,...m}，c＝{1,2,...8}，则

即字符s_j与原型S_j相匹配；其中f_pj和f_cj为步骤3中6项特性之一；if

And p={1,2,...m}, c={1,2,...8}, then

That is, the character s _j matches the prototype S _j ; where f _pj and f _cj are one of the six characteristics in step 3;

则数字串识别成功。If for the digital string C=(s ₁ ,s ₂ ,...,s ₁₂ )if there exists a prototype set P=(S ₁ ,S ₂ ,...,S _k )for

Then the identification of the number string is successful.

The digital prototype vector of this embodiment is as follows:

S ₀ ={cd=2,bp=0,rp=0,lbp=0};

S ₁ ={cd=1,wd<0.35};

S ₃ ={cd=1, lup≥1, ldp≥1};

S ₄ ={cd=2,bp≥1};

S ₅ ={cd=1,ldp≥1,lup≥1};

S ₆ ={cd=2,rp≥1};

S ₇ ={cd=1,bp≥1};

S ₈ ={cd=3};

S ₉ ={cd=2, Ibp≥1}.

Step 4: verify the identification result of the digital string;

Step 4.1: Obtain the barcode image of the online banking shield;

The barcode format used by the online banking shield is Code 128C, and its encoding method is shown in Figure 6. Each symbol in the start code, coding area and check code is composed of 3 black bars and 3 white bars, each symbol is 11 in length and contains two digits, and the end code is composed of 4 black bars and 3 white bars. The length is 13. From this, the structure of the online banking shield barcode is derived as follows:

X=[S,A ₁ ,A ₂ ,A ₃ ,A ₄ ,A ₅ ,A ₆ ,T,E]

Among them, S and E are the start and stop codes, T is the check code, and A is the character code.

Step 4.2: Use Otsu's algorithm to binarize the image and extract the barcode area, then median filter to denoise;

Step 4.3: Use the mean value method and the normalization method to eliminate the detection error caused by the tilt;

The bar code may be inclined due to deformation during packaging and transportation and the bar code pasting error. For this purpose, the average value method and the normalization method are adopted in this embodiment to eliminate the detection error caused by the inclination.

First, the width of black and white barcodes is counted line by line, and the statistical results where the total number of black and white barcodes is not 55 are removed, so as to eliminate the statistical error caused by the skew of the barcode, and the statistics are obtained:

x _i =[w ₁ ,w ₂ ,...,w ₅₅ ]i∈[1,L];

Among them, x _i represents that each row counts a total of 55 black bars and white bars to obtain the width w _k ;

Statistical process: For oblique barcodes, the number of statistical lines for some lines is less than 55 (line b), only the lines (a) whose scanning result of black and white bars is 55 are counted, and L is the total number of lines that finally meet the conditions.

Average the results to reduce statistical error:

表示每个黑条或白条统计得到的平均宽度；in,

Indicates the average width of each black bar or white bar;

Calculate barcode unit length:

The actual barcode obtained from this is:

表示每个黑条或白条归一化得到的最终二值化对应的标准宽度；in,

Indicates the standard width corresponding to the final binarization obtained by normalizing each black bar or white bar;

且

则进行编码区识别，对照Code 128C码编码表得[A₁,A₂,A₃,A₄,A₅,A₆,T]；Check the start code and end code, if

and

Then carry out coding region identification, and obtain [A ₁ , A ₂ , A ₃ , A ₄ , A ₅ , A ₆ , T] according to the Code 128C code coding table;

Finally, check the barcode:

T ^* =(105+ ₁ *A1+ ₂ *A2+ ₃ *A3+ ₄ *A4+ ₅ *A5+ ₆ *A6)%103;

If T ^* =T, the barcode recognition is successful;

If the result [A ₁ , A ₂ , A ₃ , A ₄ , A ₅ , A ₆ ] = (s ₁ , s ₂ ,..., s ₁₂ ) is obtained through the above identification, the online banking shield is qualified and the verification is passed. Otherwise, the online banking shield is incorrectly packaged or the serial number is incorrectly identified.

This embodiment identifies 100 online banking shields with a total of 1200 characters and 100 barcodes, and identifies the serial numbers [s ₁ , s ₂ ,..., s ₁₂ ] of each online banking shield and the barcodes scanned on the packaging box. No. [A ₁ , A ₂ , A ₃ , A ₄ , A ₅ , A ₆ ] for comparison, if the numbers are consistent, it is judged to be correct, otherwise it is judged that the processing and packaging is wrong or the serial number is wrongly identified, and a sound alarm is set to prompt the current The online banking shield needs to be re-checked, and at the same time, it is checked whether the number of the online banking shield is continuous.

In the experiment, the recognition accuracy rate of the algorithm reached 99%. The only error in the recognition of the online banking shield was due to the fact that a little metal dust was used to block the number 9 after processing, which made the recognition invalid. After wiping off the dust, the recognition was correct. The recognition accuracy of the online banking shield can reach 100%, and the average recognition time of the algorithm is 0.15s. Finally, the error product verification experiment is carried out. After the product is wrongly assembled with the packaging box, the system effectively detects the error and alarms.

It should be understood that the parts not described in detail in this specification belong to the prior art.

It should be understood that the above description of the preferred embodiments is relatively detailed, and therefore should not be considered as a limitation on the protection scope of the patent of the present invention. In the case of the protection scope, substitutions or deformations can also be made, which all fall within the protection scope of the present invention, and the claimed protection scope of the present invention shall be subject to the appended claims.

Claims (7)

1. A method for identifying the number of an online silver shield based on a connected domain and a tangent slope is characterized by comprising the following steps:

step 1: preprocessing the image;

step 2: dividing a digit string;

splicing broken characters and extracting adhesive characters;

the method for extracting the sticky characters specifically comprises the following substeps:

step 2.2.1: character segmentation;

obtaining upper and lower contour features through tracking search;

the contour function on the image is defined as f (X), where X ═ X₁,x₂,...,x_n)^TTo define a domain; n is the right boundary of the definition domain, namely the total point number; for X^*E R, if there is some epsilon > 0, all are compared with X^*X is less than epsilon_aAll the epsilon X satisfy inequality f (X)_a)≥f(X^*) I.e. X_aBelongs to X and | X_a-X^*If | < ε, then it is called X^*Is f (x) a local minimum on R, the local minimum is denoted as f_T(X^*) (ii) a Thereby obtaining the vector of the local minimum value of the upper profile as X_T＝(X_T1，X_T2，...，X_Tm)；

Obtaining the local maximum vector of the lower contour as X by the same method_B＝(X_B1，X_B2，...，X_Bm) (ii) a Local maximum is noted as f_B(X^*)；

For X_Ti∈X_TIf, if

And X_Bj∈[X_Ti-5,X_Ti+5]And is and

then

Dividing points for the digit strings; recording the final area height ah as a character height value;

to X_TAfter traversing, if D_y＝[D₁,D₂,...,D₁₃]The numeric string is successfully segmented, otherwise, the numeric string is failed to segment;

step 2.2.2: character segmentation;

processing the image into a standard image with uniform size, ensuring that the maximum width of all characters is W and the maximum height is H;

step 2.2.3: extracting contour mutation points;

identifying the numbers by adopting a contour slope mutation detection algorithm;

firstly, a left-right scanning method is adopted to obtain a left end outline, then a right-left scanning method is adopted to obtain a right end outline, and the abscissa of the left outline and the abscissa of the right outline are respectively recorded by using one-dimensional arrays left (x) and right (x); obtaining a bottom contour by adopting a bottom-up scanning method, and recording the vertical coordinate of the bottom contour by using a one-dimensional array bottom (i);

then calculating the slope between two adjacent points of the projection contour, judging whether the slope generates mutation, recording the frequency of the slope mutation as a character recognition characteristic, and removing the head end and the tail end of the mutation during calculation;

the slope between two adjacent points of the bottom contour is as follows:

B(i)＝(Bottom(i+1)-Bottom(i))/1；

wherein, i ═ 4,5., W-4;

when B (i) > 0.15 × H, recording that the slope generates one mutation, and recording the mutation frequency as bp;

the slope between two adjacent points of the right contour is:

R(i)＝(Right(i+1)-Right(i))/1；

wherein, i is 4,5,., H-4;

in order to improve the identification precision, the left outline is divided into an upper part and a lower part;

calculating the slope between two adjacent points of the upper left contour of the image as follows:

Lu(i)＝(Left(i+1)-Left(i))/1；

wherein, i is 4,5, H/2+ 1;

calculating the slope between two adjacent points of the lower left contour of the image as follows:

Ld(i)＝(Left(i+1)-Left(i))/1；

wherein, i ═ H/2+ 1.., H-4;

setting a slope threshold value to be 0.3W, recording the slope as one time of mutation when any one value of R (i), Lu (i) and Ld (i) is more than 0.3W, and respectively recording the number of contour mutation points in R (i), Lu (i) and Ld (i) as rp, lup and ldp;

and step 3: recognizing a numeric string;

the method comprises the steps of identification feature selection and character classification identification.

2. The method for identifying the internet banking shield number based on the connected component and the tangent slope according to claim 1, wherein the detailed implementation of the step 1 comprises the following sub-steps:

step 1.1: carrying out image binarization;

carrying out binarization processing on the image by adopting a dynamic threshold processing method Otsu's method;

step 1.2: extracting a digit string;

adopting a statistical method for the image after binarization processing, wherein 0 represents a background and 1 represents a primitive boundary; counting the number of pixels with the gray level of 1 in the horizontal direction, removing the rows with the pixels lower than the average value, extracting the region with the largest number of pixels, then carrying out pixel counting in the vertical direction on the region, removing the columns with the pixels lower than the average value, then extracting the region with the largest number of pixels, namely the region where the target digit string is located, and recording the final region height ah as a character height value.

3. The internet banking shield number identification method based on the connected component and the tangent slope as claimed in claim 1, wherein the step 2 of splicing the broken characters specifically comprises the following sub-steps:

step 2.1.1: filtering and denoising the digital string;

and (3) eliminating isolated noise points by adopting median filtering, wherein the size of a filtering window is 3 multiplied by 3, and the definition of a median filter is as follows:

G(m,n)＝MedA{fA(m+k,n+l),(k,l)∈w}；

wherein, fA (m, n) represents the original image, G (m, n) represents the processed image, m represents the number of rows, and n represents the number of columns; k e-1, l e-1, 1; w represents a two-dimensional template, 3 x 3;

step 2.1.2: removing the break point of the numeric string;

expanding by using the logarithmic string; the expansion is an operation of combining background points in contact with the primitives into the primitives to make the primitives in the image "thick", wherein A is expanded by B and is defined as:

in the formula: phi is the empty set,

is a structural element, i.e. A is BDilation is a set consisting of the location of the origin of all structural elements.

4. The internet banking shield digital identification method based on the connected component and the tangent slope according to claim 1, wherein: selecting the identification characteristics in the step 3, and distinguishing characters by adopting a numerical characteristic extraction method; for accurate classification and identification of numbers, 6 geometric and topological features are selected to construct digital prototypes in a database so as to distinguish characters, and the selected features are as follows:

(1) the number of connected domains cd is {1,2, 3 };

(2)3/4 the ratio of the maximum width within the character height to the total character height, wd, max right (i) -left (i) }/H;

(3) the number bp of bottom contour mutation points is {0 ≧ 1 };

(4) the number rp of the right contour mutation points is {0 ≧ 1 };

(5) the number of upper left contour mutation points lup ≧ 1, ≧ 0;

(6) the number of lower left contour discontinuities ldp ≧ 0, ≧ 1.

5. The method for internet banking shield digital identification based on connected component and tangent slope according to claim 4, wherein: in the step 3, the character classification and identification is to scan and segment the obtained character s_iAnd the digital prototype S in the database_iPerforming matching to perform character recognition;

digital prototype S_jSo that the number has 10 j e [0,9 ]]K is the number of features, and there are 6 features in total, but the number of features is the largest S_jTo 0, 4 features were selected, with the least number of features S_jFor 8, 1 feature is chosen, so k ∈ [1,4 ]]；

Wherein S_j＝(f_1j,f_2j,...,f_mj)，j＝{1,2,...,k}，m≤6，f_mjIs one of the 6 characteristics in the step 3; s_j＝({cd_j},{wd_j},{bp_j},{rp_j},{lup_j},{ldp_j})；

If it is not

And p ═ 1,2,. m }, and c ═ 1,2,. 6}, then

I.e. the character s_jAnd the prototype S_jMatching; wherein f is_pjAnd f_cjIs one of the 6 characteristics in the step 3;

if C is(s) for the digit string₁,s₂,…,s₁₂) If there is a prototype set P ═ (S)₁,S₂,...,S_k) For the

The digit string identification is successful.

6. The method for identifying the internet banking shield number based on the connected component and the tangent slope according to any one of claims 1 to 5, wherein: step 3, after the digital string recognition is finished, further checking the digital string recognition result;

the method specifically comprises the following substeps:

step 4.1: acquiring an online silver shield bar code image;

the online banking shield bar Code format is Code 128C Code, and the coding mode of the online banking shield bar Code format is a start Code, a coding region, a check Code and a stop Code in sequence; each symbol in the start code, the coding region and the check code consists of 3 black bars and 3 white bars, each symbol has the length of 11 and comprises two digits, and the stop code consists of 4 black bars and 3 white bars, and has the length of 13; the internet bank shield bar code structure is as follows:

X＝[S,A₁,A₂,A₃,A₄,A₅,A₆,T,E]；

wherein S, E is start code and stop code, T is check code, A_iIs character code, i is 1,2, …, 6;

step 4.2: performing binarization on the image by adopting an Otsu's algorithm, extracting a bar code region, and then performing median filtering and denoising;

step 4.3: and (3) eliminating detection errors caused by inclination by adopting an average value method and a normalization method.

7. The method for identifying the internet banking shield number based on the connected domain and the tangent slope according to claim 6, wherein the specific implementation process of the step 4.3 is as follows:

firstly, the width of the black and white bar code is counted line by line, and the counting result that the total number of the black and white bar code is not 55 is removed, thereby eliminating the counting error caused by the inclination of the bar code, and obtaining the following result by counting:

x_i＝[w₁,w₂,...,w₅₅]i∈[1,L]；

wherein x is_iRepresenting a total of 55 black and white bars counted per line for a width w_k(ii) a L is the total number of rows which finally meet the condition;

the results are averaged to reduce statistical error:

wherein,

representing the average width obtained by each black strip or white strip;

calculating the unit length of the bar code:

the actual bar code is thus obtained as:

wherein,

representing each black bar or whiteNormalizing the bars to obtain a final standard width corresponding to binarization;

checking the start code and the end code if

And is

Then the Code area identification is carried out, and the Code 128C Code table is compared to obtain [ A ]₁,A₂,A₃,A₄,A₅,A₆,T]；

And finally, checking the bar code:

T^*＝(105+1*A₁+2*A₂+3*A₃+4*A₄+5*A₅+6*A₆)％103；

if T^*If the bar code is T, the bar code identification is successful;

if the result [ A ] is obtained by the above recognition₁,A₂,A₃,A₄,A₅,A₆]＝(s₁,s₂,...,s₁₂) The online silver shield is qualified, the verification is passed, otherwise, the online silver shield is packaged or the serial number identification is mistaken.

Images