Facial vein image identity recognition method based on cutting depth

Abstract

The present application relates to a kind of face vein image identity recognition methods based on cutting depth, belong to biological recognition technical field, comprising: face vein image is divided into several equal size sub-region images, different cutting scales are sequentially given to each sub-region image to obtain the scale space of sub-region image, the cutting depth of sub-region image is calculated based on the scale space of sub-region image;The cutting depth of all sub-region images is weighted and fused to obtain the cutting depth of face vein image;According to the cutting depth of face vein image, feature coding is carried out to face vein image, and feature coding image is formed;According to feature coding image, identity recognition is carried out to face vein image.The method of the present application can accurately realize identity authentication by the face vein image of user.

Description

Technical Field

[0001] This invention relates to the field of biometrics, and in particular to a method for facial vein image identification based on cutting depth. Background Technology

[0002] In recent years, with the development of information technology, facial recognition has been widely used in daily life, bringing great convenience. Existing facial recognition methods generally rely on images captured by RGB cameras. However, because the face is an external feature, it is easily copied from photos, screens, headgear, etc., making liveness detection a major challenge. Furthermore, twins have extremely similar facial features, which also poses a significant challenge. Vein recognition, as a second-generation biometric technology, possesses advantages such as being an internal feature, uncopyable, having a high level of security, and accommodating significant individual differences, making it of great research value. Currently, finger vein recognition, as an emerging identification technology, has been applied in fields such as finance and military, but methods based on facial vein recognition are still rarely studied. Summary of the Invention

[0003] The main objective of this invention is to provide a facial vein image identification method based on cutting depth. This method extracts features from facial vein images based on cutting depth, more accurately expresses the changes between vein details, and facilitates identification.

[0004] To achieve the above objectives, the present invention provides a facial vein image identification method based on cutting depth, comprising the following steps:

[0005] Step 1. Divide the facial vein image into several sub-region images of equal size, assign different cutting scales to each sub-region image to obtain the scale space of the sub-region image, and calculate the cutting depth of the sub-region image based on the scale space of the sub-region image.

[0006] Step 2. Weighted fusion of the cutting depths of all sub-region images to obtain the cutting depth of the facial vein image;

[0007] Step 3. Encode the facial vein image based on the cutting depth to form a feature-coded image;

[0008] Step 4. Identify the facial vein image based on the feature-encoded image.

[0009] Preferably, calculating the cutting depth of the sub-region image in step 1 specifically includes:

[0010] Step 1.1. Divide the facial vein image into several sub-region images of equal size, wherein the number of rows in each sub-region image is P and the number of columns is Q;

[0011] Step 1.2. Assign different cutting scales to each sub-region image to obtain the scale space L(i, j, θ) of the sub-region image. The calculation formula is:

[0012]

[0013]

[0014] In the formula, I(i,j) represents the sub-region image, i is the row coordinate, j is the column coordinate, θ is the cutting scale, and G is a two-dimensional Gaussian function.

[0015] Step 1.3. Construct a (2n+1)*(2n+1) detection box on the sub-region image, and calculate the cutting depth of the center pixel within the detection box region. The calculation formula is as follows:

[0016]

[0017] In the formula, C(i, j, θ) is the cutting depth of pixel (i, j), and h min H represents the minimum grayscale value within the detection bounding box area. min The minimum gray value within the sub-region image; m x (i,j) represents the average gray level within the detection area; M x (i, j, θ) represents the average gray level of the sub-region image; Let c be the variance within the detection area; c is a constant with a value range of [0, 1].

[0018] Step 1.4. Traverse the detection box across the sub-region image to complete the calculation of the cutting depth of the sub-region image.

[0019] Preferably, the average grayscale value M of the sub-region image x The formula for calculating (i, j, θ) is:

[0020]

[0021] average gray level m within the detection area x The formula for calculating (i,j) is:

[0022]

[0023] In the formula, x(a,b) is the gray value of pixel (a,b) within the detection region;

[0024] Variance within the detection area The calculation formula is:

[0025]

[0026] Preferably, in step 2, the cutting depths of all sub-region images are weighted and fused to obtain the cutting depth S of the facial vein image N (i,j) is calculated by the formula:

[0027]

[0028]

[0029] In the formula, n is the digital index of the sub-region image, w n is the weight of the n-th sub-region image, and N represents the number of sub-regions.

[0030] Preferably, in step 3, performing feature encoding on the facial vein image according to the cutting depth of the facial vein image specifically comprises: setting a cutting depth threshold Tn;

[0031] If S N (i,j)<(Tn-30), then the feature code of this pixel point is set to 0, that is, X(i,j)=0;

[0032] If (Tn-30)<S N (i,j)<(Tn-20), then the feature code of this pixel point is set to 1, that is, X(i,j)=1;

[0033] If (Tn-20)<S N (i,j)<(Tn-10), then the feature code of this pixel point is set to 2, that is, X(i,j)=2;

[0034] If (Tn-10)<S N (i,j)<Tn, then the feature code of this pixel point is set to 3, that is, X(i,j)=3;

[0035] If Tn<S N (i,j)<(Tn+10), then the feature code of this pixel point is set to 4, that is, X(i,j)=4;

[0036] If (Tn+10)<S N (i,j)<(Tn+20), then the feature code of this pixel point is set to 5, that is, X(i,j)=5;

[0037] If (Tn+20)<S N (i,j)<(Tn+30), then the feature code of this pixel point is set to 6, that is, X(i,j)=6;

[0038] If (Tn+30)<S N (i,j), then the feature code of this pixel point is set to 7, that is, X(i,j)=7.

[0039] Preferably, step 4 involves identifying the facial vein image based on feature encoding:

[0040] Step 4.1. Obtain the feature encoding of the facial vein image of the user to be identified;

[0041] Step 4.2. Calculate the angular distance D(U,V) between the feature code of the facial vein image of the user to be identified and the template feature code, and perform identity recognition based on the angular distance. The formula for calculating the angular distance is:

[0042]

[0043]

[0044] In the formula, U is the feature-coded image of the user to be identified, V is the template feature-coded image, p is the number of rows in the feature-coded image of the user to be identified and the template feature-coded image, and q is the number of columns in the feature-coded image of the user to be identified and the template feature-coded image.

[0045] Preferably, before dividing the facial vein image into several equally sized sub-region images in step 1, the process includes enhancing the facial vein image, specifically including:

[0046] Step a. Perform a filtering operation on the facial vein image f(x,y) to obtain the filtered facial vein image F(x,y);

[0047] Step b. Map the facial vein image f(x,y) based on the logarithm of the average gray value to obtain the processed mapped image f. bri (x,y);

[0048] Step c. Map the image f bri The illumination component L(x,y) is obtained by performing mask calculations using (x,y) and the normalized center wrap function g(x,y);

[0049] Step d. Subtract the illumination component L(x,y) from the filtered facial vein image F(x,y) to obtain the reflection component K(x,y);

[0050] Step e. Perform an exponential transform on the reflection component K(x,y) to obtain the enhanced facial vein image Z(x,y);

[0051] Step f. Quantize the image Z(x,y) into a gray value in the range of 0-255 to obtain the target enhanced vein image B(x,y).

[0052] Preferably, the formula for calculating F(x,y) is:

[0053]

[0054]

[0055] In the formula, δ r With δ d For smoothing parameters, f(i,j) represents the gray value of pixel (i,j) within the filtering window, x represents the row coordinate of the pixel, and y represents the column coordinate of the pixel;

[0056] The mapped image f bri The formula for calculating (x,y) is:

[0057]

[0058]

[0059] In the formula, Let f be a constant, ranging from [0.001, 0.1], where S represents the number of rows in the facial vein image, T represents the number of columns in the facial vein image, and f is a constant. max (x,y) represents the maximum gray value of the facial vein image f(x,y). The logarithmic mean gray value of the facial vein image f(x,y) is represented.

[0060] The formula for calculating the irradiation component L(x,y) is as follows:

[0061] L(x,y)=f bri (x,y)×g(x,y)

[0062]

[0063] In the formula, A is the normalization constant, and μ is the function wrapping scale;

[0064] The formula for calculating the reflection component K(x,y) is as follows:

[0065] K(x,y)=ln[F(x,y)]-ln[L(x,y)]

[0066] Substituting the formula for calculating L(x,y) into the formula for calculating K(x,y), we get:

[0067]

[0068] Preferably, the formula for calculating the enhanced facial vein image Z(x,y) is:

[0069]

[0070] The formula for calculating the target enhanced vein image B(x,y) is as follows:

[0071]

[0072] In the formula, Z max and Z min These represent the maximum and minimum gray values ​​of the enhanced facial vein image Z(x,y), respectively.

[0073] Compared with existing technologies, this invention has the following technical advantages: Because the grayscale level of the facial vein region is low while the background grayscale level is relatively high, it exhibits a valley-shaped characteristic in the longitudinal section. This invention incorporates cutting depth into facial vein image feature extraction. Facial vein features extracted using this method can more accurately express the changes between vein details, thus making the facial vein feature description more accurate. Furthermore, this invention uses a sliding window thresholding method to encode the surface cutting depth as a feature, and then uses angular distance to calculate the feature similarity between two facial vein images, improving the accuracy of facial vein recognition. Attached Figure Description

[0074] Figure 1 This is a flowchart of the facial vein image identification method based on cutting depth that relates to the present invention. Detailed Implementation

[0075] To make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be specifically described below in conjunction with the embodiments and accompanying drawings, but the scope of protection of the present invention is not limited thereto.

[0076] See attached document Figure 1 As shown, this invention relates to a facial vein image identification method based on cutting depth, specifically including the following steps:

[0077] Step 1. Divide the facial vein image into several equally sized sub-region images. Assign different cutting scales to each sub-region image to obtain the scale space of the sub-region image. Calculate the cutting depth of the sub-region image based on the scale space of the sub-region image. The specific steps are as follows:

[0078] Step 1.1. Divide the facial vein image into several sub-region images of equal size, wherein the number of rows in each sub-region image is P and the number of columns is Q;

[0079] Step 1.2. Assign different cutting scales to each sub-region image to obtain the scale space L(i,j,θ) of the sub-region image. The calculation formula is:

[0080]

[0081]

[0082] In the formula, I(i,j) represents the sub-region image, i is the row coordinate, j is the column coordinate, θ is the cutting scale, and G is a two-dimensional Gaussian function.

[0083] Step 1.3. Construct a (2n+1)*(2n+1) detection box on the sub-region image, and calculate the cutting depth of the center pixel within the detection box region. The calculation formula is:

[0084]

[0085] In the formula, C(i,j,θ) is the cutting depth of pixel (i,j), and h min H represents the minimum grayscale value within the detection bounding box area. min The minimum gray value within the sub-region image; m x (i,j) represents the average gray level within the detection area; M x (i,j,θ) represents the average gray level of the sub-region image; Let c be the variance within the detection area; c is a constant with a value range of [0, 1].

[0086] Wherein, the average gray level M of the sub-region image x The formula for calculating (i,j,θ) is:

[0087]

[0088] average gray level m within the detection area x The formula for calculating (i,j) is:

[0089]

[0090] In the formula, x(a,b) is the gray value of pixel (a,b) within the detection region;

[0091] Variance within the detection area The calculation formula is:

[0092]

[0093] Step 1.4. Traverse the detection box across the sub-region image to complete the calculation of the cutting depth of the sub-region image.

[0094] Step 2. Weighted fusion of the cutting depths of all sub-region images to obtain the cutting depth S of the facial vein image. N (i,j), the calculation formula is:

[0095]

[0096]

[0097] In the formula, n is a digital index of the sub-region image, w n is the weight of the n-th sub-region image, and N represents the number of sub-regions.

[0098] Step 3. Perform feature encoding on the facial vein image according to the cutting depth of the facial vein image to form a feature encoded image, and the specific steps are: setting a cutting depth threshold Tn;

[0099] If S N (i,j) < (Tn-30), then the feature code of this pixel is set to 0, that is, X(i,j)=0;

[0100] If (Tn-30) < S N (i,j) < (Tn-20), then the feature code of this pixel is set to 1, that is, X(i,j)=1;

[0101] If (Tn-20) < S N (i,j) < (Tn-10), then the feature code of this pixel is set to 2, that is, X(i,j)=2;

[0102] If (Tn-10) < S N (i,j) < Tn, then the feature code of this pixel is set to 3, that is, X(i,j)=3;

[0103] If Tn < S N (i,j) < (Tn+10), then the feature code of this pixel is set to 4, that is, X(i,j)=4;

[0104] If (Tn+10) < S N (i,j) < (Tn+20), then the feature code of this pixel is set to 5, that is, X(i,j)=5;

[0105] If (Tn+20) < S N (i,j) < (Tn+30), then the feature code of this pixel is set to 6, that is, X(i,j)=6;

[0106] If (Tn+30) < S N (i,j), then the feature code of this pixel is set to 7, that is, X(i,j)=7.

[0107] Step 4. Perform identity recognition on the facial vein image according to the feature encoded image, and the specific steps are:

[0108] Step 4.1. Obtain the feature code of the facial vein image of the user to be identified;

[0109] Step 4.2. Calculate the angular distance D(U,V) between the feature code of the facial vein image of the user to be identified and the template feature code, and perform identity recognition based on the angular distance. The formula for calculating the angular distance is:

[0110]

[0111]

[0112] In the formula, U is the feature-coded image of the user to be identified, V is the template feature-coded image, p is the number of rows in the feature-coded image of the user to be identified and the template feature-coded image, and q is the number of columns in the feature-coded image of the user to be identified and the template feature-coded image.

[0113] In one embodiment, before dividing the facial vein image into several equally sized sub-region images, the process includes enhancing the facial vein image. The specific steps are as follows:

[0114] Step a. Filter the facial vein image f(x,y) to obtain the filtered facial vein image F(x,y). The calculation formula is:

[0115]

[0116]

[0117] In the formula, δ r With δ d For smoothing parameters, f(i,j) represents the gray value of pixel (i,j) within the filtering window, x represents the row coordinate of the pixel, and y represents the column coordinate of the pixel;

[0118] Step b. Map the facial vein image f(x,y) based on the logarithm of the average gray value to obtain the processed mapped image f. bri (x, y), the calculation formula is:

[0119]

[0120]

[0121] In the formula, Let f be a constant, ranging from [0.001, 0.1], where S represents the number of rows in the facial vein image, T represents the number of columns in the facial vein image, and f is a constant. max (x,y) represents the maximum gray value of the facial vein image f(x,y). The logarithmic mean gray value of the facial vein image f(x,y) is represented.

[0122] Step c. Map the image f briThe illumination component L(x,y) is obtained by masking the irradiation components L(x,y) using the normalized center wrap function g(x,y) and the formula is as follows:

[0123] L(x,y)=f bri (x,y)×g(x,y)

[0124]

[0125] In the formula, A is the normalization constant, and μ is the function wrapping scale;

[0126] Step d. Subtract the illumination component L(x,y) from the filtered facial vein image F(x,y) to obtain the reflection component K(x,y). The calculation formula is as follows:

[0127] K(x,y)=ln[F(x,y)]-ln[L(x,y)]

[0128] Substituting the formula for calculating L(x,y) into the formula for calculating K(x,y), we get:

[0129]

[0130] Step e. Perform an exponential transform on the reflection component K(x,y) to obtain the enhanced facial vein image Z(x,y). The calculation formula is as follows:

[0131]

[0132] Step f. Quantize the image Z(x,y) into a grayscale value in the range of 0-255 to obtain the target enhanced vein image B(x,y). The calculation formula is as follows:

[0133]

[0134] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A facial vein image identification method based on cutting depth, characterized in that... Includes the following steps: Step 1. Divide the facial vein image into several equally sized sub-region images. Assign different cutting scales to each sub-region image to obtain its scale space. Calculate the cutting depth of the sub-region image based on its scale space. Specifically, this includes: Step 1.

1. Divide the facial vein image into several sub-region images of equal size, wherein the number of rows in each sub-region image is P and the number of columns is Q; Step 1.

2. Assign different cutting scales to each sub-region image to obtain the scale space of the sub-region image. The calculation formula is: In the formula, I(i,j) represents the sub-region image, where i is the row coordinate and j is the column coordinate. Let G be the cutting scale, and G be a two-dimensional Gaussian function. Step 1.

3. Construct a (2n+1)×(2n+1) detection box on the sub-region image, and calculate the cutting depth of the center pixel within the detection box region. The calculation formula is as follows: In the formula, Let (i, j) be the cutting depth. The minimum grayscale value within the detection box area; The minimum gray value within the sub-region image; This represents the average grayscale value within the detection area. This represents the average grayscale value of the sub-region image; The variance within the detection area; It is a constant whose value range is [0,1]; Step 1.

4. Traverse the detection box across the sub-region image to calculate the cutting depth of the sub-region image; Step 2. Weighted fusion of the cutting depths of all sub-region images to obtain the cutting depth of the facial vein image; Step 3. Encode the facial vein image based on the cutting depth to form a feature-coded image; Step 4. Identify the facial vein image based on the feature-encoded image.

2. The facial vein image identification method based on cutting depth according to claim 1, characterized in that, The average grayscale value of the sub-region image The calculation formula is: Average gray level within the detection area The calculation formula is: In the formula, To detect pixels within the region grayscale value; Variance within the detection area The calculation formula is: 。 3. The facial vein image identification method based on cutting depth according to claim 1, characterized in that, Step 2 involves weighted fusion of the cutting depths of all sub-region images to obtain the cutting depth of the facial vein image. The calculation formula is: In the formula, n is the numerical index of the sub-region image. It is the weight of the nth sub-region image, where N represents the number of sub-regions.

4. The facial vein image identification method based on cutting depth according to claim 3, characterized in that, Step 3 involves feature encoding of the facial vein image based on the cutting depth of the facial vein image, specifically including: setting a cutting depth threshold. ; like If so, then the feature code of that pixel is set to 0, that is... ; like Then the feature code of that pixel is set to 1, that is... ; like Then the feature code of that pixel is set to 2, that is... ; like Then the feature code of that pixel is set to 3, that is... ; like Then the feature code of that pixel is set to 4, that is... ; like Then the feature code of that pixel is set to 5, that is... ; like Then the feature code of that pixel is set to 6, that is... ; like Then the feature code of that pixel is set to 7, that is... .

5. The facial vein image identification method based on cutting depth according to claim 1, characterized in that, Step 4, which involves identifying the facial vein image based on feature encoding, specifically includes: Step 4.

1. Obtain the feature encoding of the facial vein image of the user to be identified; Step 4.

2. Calculate the angular distance between the feature encoding of the facial vein image of the user to be identified and the feature encoding of the template. The identity is identified based on the angular distance, and the formula for calculating the angular distance is: In the formula, U is the feature-coded image of the user to be identified, V is the template feature-coded image, p is the number of rows in the feature-coded image of the user to be identified and the template feature-coded image, and q is the number of columns in the feature-coded image of the user to be identified and the template feature-coded image.

6. The facial vein image identification method based on cutting depth according to claim 1, characterized in that, Before dividing the facial vein image into several equally sized sub-region images in step 1, the facial vein image undergoes enhancement processing, specifically including: Step a. Image of facial veins Perform filtering to obtain the filtered facial vein image. ; Step b. Image of facial veins The mapped image is obtained by performing a mapping operation based on the logarithm of the average gray value. ; Step c. Map the image and normalized center-around function Perform mask calculations to obtain the irradiation component. ; Step d. Filter the facial vein image Subtract the irradiation component The reflection component is obtained. ; Step e. For the reflection component Then, an exponential transformation is performed to obtain the enhanced facial vein image. ; Step f. Transfer the image The target enhanced vein image is obtained by quantizing to a grayscale value in the range of 0-255. .

7. The facial vein image identification method based on cutting depth according to claim 6, characterized in that, The The calculation formula is: In the formula, and For smoothing parameters, Represents the number of pixels within the filtering window The grayscale value, where x represents the row coordinate of the pixel and y represents the column coordinate of the pixel; The mapped image The calculation formula is: In the formula, Let S be a constant, ranging from [0.001, 0.1], where S represents the number of rows in the facial vein image and T represents the number of columns in the facial vein image. Image of facial veins Maximum grayscale value, Image of facial veins The logarithmic mean gray value; The irradiation component The calculation formula is: In the formula, A is the normalization constant. The function wraps around the scale; The reflection component The calculation formula is: Formula Substitute the calculation formula into The calculation formula yields: 。 8. The facial vein image identification method based on cutting depth according to claim 7, characterized in that, The enhanced facial vein image The calculation formula is: The target enhanced vein image The calculation formula is: In the formula, and These represent enhanced facial vein images. The maximum and minimum grayscale values.

Images