A detection and processing method of a hair in a dorsal digital vein image recognition process

Abstract

The present application relates to a kind of detection and processing method of hair in dorsal digital vein image recognition process, belong to the field of biometric identification, including the following steps: collecting the dorsal digital vein image of user;Calculate the curvature value of each pixel point in dorsal digital vein image cross-symmetry direction, according to the curvature value of cross-symmetry direction, filter pending hair point and construct first hair binary image;Set connected domain detection threshold, the area of all connected domain of first hair binary image is counted, if the area of the connected domain is greater than connected domain detection threshold, then the connected region is excluded, obtain second hair binary image;Through second hair binary image, the dorsal digital vein image of repair is obtained to remove hair;The dorsal digital vein image of removing hair is carried out image enhancement processing.The present application can improve the security and stability of dorsal digital vein recognition algorithm.

Description

Technical Field

[0001] This invention relates to the field of biometrics, and in particular to a method for detecting and processing hair during the recognition of dorsal finger vein images. Background Technology

[0002] Finger vein recognition technology uses near-infrared light to illuminate the back of the finger. When the hemoglobin in the blood flowing in the finger absorbs the near-infrared light, an image of the finger vein can be obtained. Then, a feature extraction algorithm is used to extract feature values ​​from the finger vein image. During recognition, the extracted feature value data is compared with the previously stored feature value data to provide the recognition comparison result, thereby achieving the function of identity recognition.

[0003] Most finger vein recognition technologies on the market use finger pad vein recognition technology. However, in recent years, finger dorsal vein recognition technology has attracted market attention due to its unique stability. For example, when using finger dorsal vein recognition technology, pressing the authentication device with a finger does not affect the blood vessels and blood flow in the finger dorsal vein, nor does lifting heavy objects affect the blood vessels and blood flow in the finger dorsal vein. Therefore, it does not affect the recognition performance of biometrics using finger dorsal veins. This can solve the problem of low verification success rate of vein recognition authentication caused by pressing the finger with a finger pad and lifting heavy objects when using finger pad vein recognition technology.

[0004] During the development of digital dorsal vein technology, it was found that some users, especially male users, have particularly dense hair on the back of their fingers, which affects the feature extraction and recognition performance of digital dorsal veins. Therefore, it is necessary to develop a new method to eliminate the influence of hair on digital dorsal vein recognition technology. Summary of the Invention

[0005] The main objective of this invention is to provide a method for detecting and processing hair during the process of dorsal finger vein image recognition. This method can effectively filter out hair and then repair the hair area, reducing the impact of the hair area on dorsal finger vein recognition and greatly improving the security and stability of the dorsal finger vein recognition algorithm.

[0006] To achieve the above objectives, the present invention provides the following solution:

[0007] This invention relates to a method for detecting and processing body hair during digital dorsal vein image recognition, comprising the following steps:

[0008] S1. Acquire images of the user's dorsal digital veins;

[0009] S2. Calculate the curvature value of the cross-symmetry direction of each pixel in the dorsal vein image, filter the undetermined hair points according to the curvature value of the cross-symmetry direction, and construct the first hair binary image.

[0010] S3. Set a connected component detection threshold, calculate the area of ​​all connected components in the first hair binary graph, and if the area of ​​the connected component is greater than the connected component detection threshold, then filter out the connected component to obtain the second hair binary graph.

[0011] S4. The dorsal digital vein image is repaired by the second hair binary image to obtain the dorsal digital vein image with hair removed;

[0012] S5. Perform image enhancement processing on the digital dorsal vein image after hair removal.

[0013] Preferably, the curvature value of the cross-symmetric direction includes the curvature values ​​of the pixel in multiple directions at equal intervals and the curvature value in the 90-degree direction.

[0014] Preferably, the equally spaced angles are 45 degrees.

[0015] Preferably, step S2, which involves filtering the undetermined hair points based on multi-directional curvature values ​​and constructing a first binary image of the hair, includes:

[0016] S2.1. Calculate the curvature values ​​of each pixel in the dorsal digital vein image at equal intervals in multiple directions and the difference between the curvature values ​​in the 90-degree direction;

[0017] S2.2. Select the direction with the largest absolute value of the difference as the main direction of the pixel;

[0018] S2.3. Set a curvature threshold. If the curvature value of the pixel in the main direction is greater than the curvature threshold, then the pixel is a hair point to be determined and its pixel value is set to 1. Otherwise, it is set to 0.

[0019] Preferably, the formula for constructing the first binary image of body hair in step S2 is:

[0020] g(i) = |(C(i) - C(i + 90))|

[0021] g(a) = max(g(i))

[0022]

[0023] In the formula, C(i) is the curvature value of any pixel in the dorsal vein image at the i-angle direction; C(i+90) is the curvature value of any pixel in the dorsal vein image at the i+90 direction; g(i) is the absolute value of the difference between the curvature of any pixel in the dorsal vein image at the i-angle direction and the i+90 direction; a is the principal direction of any pixel in the dorsal vein image; z(a) is the curvature value of any pixel in the dorsal vein image at the principal direction; x is the abscissa of the hair point to be determined; y is the ordinate of the hair point to be determined; k is the curvature threshold; and I(x,y) is the first binary image of the hair.

[0024] Preferably, in step S3, if the area of ​​the connected region is greater than the connected region detection threshold, the connected region is filtered out to obtain the second hair binary image: if the area of ​​the connected region is greater than the connected region detection threshold, the pixel value of the connected region is set to 0 to obtain the second hair binary image.

[0025] Preferably, in step S4, the digital dorsal vein image obtained by repairing the digital dorsal vein image using the second hair binary image is as follows:

[0026] S4.1. Identify the hair points in the dorsal vein image based on the second hair binary image, wherein the hair points are the pixels in the dorsal vein image corresponding to the pixels with a pixel value of 1 in the second hair binary image;

[0027] S4.2. Calculate the center line of each hair and repair the hair using the non-hair area perpendicular to the center line.

[0028] Preferably, in step S4.2, the non-hair region perpendicular to the center line of the hair includes an upper non-hair region and a lower non-hair region, and the size of the upper non-hair region and the lower non-hair region is N times the radius of curvature of the hair point.

[0029] Preferably, in step S4.2, the repair of each hair point using the non-hair region perpendicular to the hair center line is as follows: grayscale values ​​of pixels in the upper and lower non-hair regions are fitted as the pixel values ​​of the hair point, wherein the fitting weight is inversely proportional to the distance of the pixel from the hair center line, and the fitting formula is:

[0030]

[0031]

[0032] In the formula, Y(n) is the pixel restoration value of the hair point at a distance n from the center line of the hair; R is the radius of curvature of the hair point; α is the fitting weight of the upper non-hair region; β is the fitting weight of the lower non-hair region; m is the pixel position of the non-hair region; I0(m) is the gray value of the pixel in the non-hair region above the center line of the hair; I1(m) is the gray value of the pixel in the non-hair region below the center line of the hair.

[0033] Preferably, the image enhancement processing of the finger dorsal vein image after hair removal in step S5 includes first performing small-scale filtering processing on the finger dorsal vein image after hair removal, and then performing large-scale filtering processing.

[0034] The present invention has the following technical effects:

[0035] 1. This invention initially locates the hair region by using the curvature value of the cross-symmetric direction. Then, it performs a second screening of the hair region by the size of the connected region, which can accurately locate the location information of the hair region. Then, based on the hair location result, the hair region is repaired to obtain the dorsal finger vein image after hair removal, thus preparing to locate the hair region for subsequent processing.

[0036] 2. This invention enhances the repaired vein image with an algorithm to further eliminate the influence of hair on vein image recognition. First, this invention performs small-scale smoothing on the finger dorsal vein image after hair removal, mainly to reduce image noise. Second, it performs large-scale smoothing on the denoised image and uses the smoothing results of two different windows to enhance the finger dorsal vein image and improve the quality of the finger dorsal vein image. Attached Figure Description

[0037] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This invention relates to a flowchart of a method for detecting and processing body hair during digital dorsal vein image recognition. Detailed Implementation

[0039] 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.

[0040] Reference Figure 1 As shown, this invention relates to a method for detecting and processing body hairs during digital dorsal vein image recognition, which includes the following steps:

[0041] S1. Collect images of the user's finger veins. Before proceeding with the next detection step, the images generally need to be normalized in size.

[0042] S2. Calculate the curvature value of the cross-symmetry direction of each pixel in the dorsal vein image. Based on the curvature value of the cross-symmetry direction, filter the undetermined hair points and construct the first hair binary image. The curvature value of the cross-symmetry direction includes the curvature values ​​of multiple directions at equal intervals and the curvature value of the 90-degree direction of the pixel. The specific steps are as follows:

[0043] S2.1. Calculate the curvature values ​​of each pixel in the dorsal digital vein image in multiple directions at equal intervals and the difference between the curvature values ​​in the 90-degree direction. The equal intervals are preferably 45 degrees.

[0044] S2.2. Select the direction with the largest absolute value of the difference as the main direction of the pixel;

[0045] S2.3. Set a curvature threshold. If the curvature value of a pixel in its principal direction is greater than the curvature threshold, then the pixel is designated as a hair point and its pixel value is set to 1; otherwise, it is set to 0. The calculation formula is:

[0046] g(i) = |(C(i) - C(i + 90))|

[0047] g(a) = max(g(i))

[0048]

[0049] In the formula, C(i) is the curvature value of any pixel in the dorsal vein image at the i-angle direction; C(i+90) is the curvature value of any pixel in the dorsal vein image at the i+90 direction; g(i) is the absolute value of the difference between the curvature of any pixel in the dorsal vein image at the i-angle direction and the i+90 direction; a is the principal direction of any pixel in the dorsal vein image; z(a) is the curvature value of any pixel in the dorsal vein image at the principal direction; x is the abscissa of the hair point to be determined; y is the ordinate of the hair point to be determined; k is the curvature threshold; and I(x,y) is the first binary image of the hair.

[0050] If the equally spaced angles are θ, then curvature values ​​in 360 / θ directions need to be calculated. That is, if the equally spaced angles are 30 degrees, then curvature values ​​in 12 directions need to be calculated; if the equally spaced angles are 45 degrees, then curvature values ​​in 8 directions need to be calculated. There is no limitation here, but 45 degrees is generally preferred. In addition, the direction with the largest curvature difference indicates that its grayscale change is the most obvious and best reflects the directionality of the pixel, so it is taken as the main direction.

[0051] Because of the significant color difference between the hair follicles and the back of the finger, they possess inherent physical characteristics. Furthermore, based on batch data analysis of finger vein images containing hair follicles, the curvature values ​​of these hair follicles fluctuate within a certain range. Therefore, a curvature threshold can be set for initial hair follicle identification. After identifying the hair follicles, their location information needs to be stored for accurate retrieval later. In this solution, a binary image is constructed to store this location information.

[0052] S3. Set a connected component detection threshold, and calculate the area of ​​all connected components in the first hair binary image. If the area of ​​the connected component is greater than the connected component detection threshold, then filter out the connected component to obtain the second hair binary image. Specifically, if the area of ​​the connected component is greater than the connected component detection threshold, set the pixel value of the connected component to 0 to obtain the second hair binary image.

[0053] Since the size of body hair is usually within a certain range, a secondary judgment can be made by statistically analyzing the size of the body hair region to remove some false body hair points. The connected component detection threshold is confirmed by batch processing of images.

[0054] S4. The dorsal digital vein image is repaired using the second hair binary image to obtain a hair-removed dorsal digital vein image, including:

[0055] S4.1. Identify the hair points in the dorsal vein image based on the second hair binary image, wherein the hair points are the pixels in the dorsal vein image corresponding to the pixels with a pixel value of 1 in the second hair binary image;

[0056] S4.2. Calculate the centerline of each hair, and repair the hair using the non-hair region perpendicular to the centerline. This non-hair region includes an upper non-hair region and a lower non-hair region, the size of which is N times the radius of curvature of the hair point. Repairing each hair point using the non-hair region perpendicular to the centerline involves fitting grayscale values ​​of pixels in the upper and lower non-hair regions as the pixel value of the hair point. The fitting weight is inversely proportional to the distance from the pixel to the hair centerline, and the fitting formula is:

[0057]

[0058]

[0059] In the formula, Y(n) is the pixel restoration value of the hair point at a distance n from the center line of the hair; R is the radius of curvature of the hair point; α is the fitting weight of the upper non-hair region; β is the fitting weight of the lower non-hair region; m is the pixel position of the non-hair region; I0(m) is the gray value of the pixel in the non-hair region above the center line of the hair; I1(m) is the gray value of the pixel in the non-hair region below the center line of the hair.

[0060] S5. Image enhancement processing is performed on the dorsal finger vein image after hair removal. Specifically, the image is first subjected to small-scale filtering, followed by large-scale filtering. In this invention, the image is first filtered at a small scale, such as by mean filtering, primarily to reduce image noise. Then, a large-scale filter is applied to the denoised image. The dorsal finger vein image is enhanced using the smoothing results from two different windows, which significantly improves the quality of the dorsal finger vein image and facilitates subsequent processing.

[0061] 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 method for detecting and processing body hair during digital dorsal vein image recognition, characterized in that: Includes the following steps: S1. Acquire images of the user's dorsal digital veins; S2. Calculate the curvature value of the cross-symmetry direction of each pixel in the dorsal vein image. The curvature value of the cross-symmetry direction includes the curvature value of the pixel in multiple directions at equal intervals and the curvature value of the 90-degree direction. Select the hair points to be determined based on the curvature value of the cross-symmetry direction and construct the first hair binary image. S3. Set a connected component detection threshold, calculate the area of ​​all connected components in the first hair binary graph, and if the area of ​​the connected component is greater than the connected component detection threshold, then filter out the connected component to obtain the second hair binary graph. S4. The dorsal digital vein image is repaired by the second hair binary image to obtain the dorsal digital vein image with hair removed; S5. Perform image enhancement processing on the digital dorsal vein image after hair removal.

2. The method for detecting and processing body hair during digital dorsal vein image recognition according to claim 1, characterized in that: Step S2, which involves filtering undetermined hair points based on multi-directional curvature values ​​and constructing a first binary image of hair, includes: S2.

1. Calculate the curvature values ​​of each pixel in the dorsal digital vein image at multiple directions with equal intervals and the difference between the curvature values ​​in the 90-degree direction; S2.

2. Select the direction with the largest absolute value of the difference as the main direction of the pixel; S2.

3. Set a curvature threshold. If the curvature value of the pixel in the main direction is greater than the curvature threshold, then the pixel is a hair point to be determined and its pixel value is set to 1. Otherwise, it is set to 0.

3. The method for detecting and processing body hair during the digital dorsal vein image recognition process according to claim 2, characterized in that: The formula for constructing the first binary image of body hair in step S2 is as follows: g(i) = |(C(i) - C(i + 90))| g(a) = max(g(i)) In the formula, C(i) is the curvature value of any pixel in the dorsal vein image at the i-angle direction; C(i+90) is the curvature value of any pixel in the dorsal vein image at the i+90 direction; g(i) is the absolute value of the curvature difference between any pixel in the dorsal vein image at the i-angle direction and the i+90 direction; a is the principal direction of any pixel in the dorsal vein image; x is the abscissa of the hair point to be determined; y is the ordinate of the hair point to be determined; k is the curvature threshold; and I(x,y) is the first binary image of the hair.

4. The method for detecting and processing body hair during digital dorsal vein image recognition according to claim 1, characterized in that: In step S3, if the area of ​​the connected component is greater than the connected component detection threshold, the connected component is filtered out to obtain the second hair binary image. If the area of ​​the connected component is greater than the connected component detection threshold, the pixel value of the connected component is set to 0 to obtain the second hair binary image.

5. The method for detecting and processing body hair during the digital dorsal vein image recognition process according to claim 2, characterized in that: In step S4, the dorsal digital vein image is repaired using the second hair binary image to obtain a hair-free dorsal digital vein image: S4.

1. Identify the hair points in the dorsal vein image based on the second hair binary image, wherein the hair points are the pixels in the dorsal vein image corresponding to the pixels with a pixel value of 1 in the second hair binary image; S4.

2. Calculate the center line of each hair, and repair each hair point by using the non-hair area perpendicular to the center line of the hair in sequence.

6. The method for detecting and processing body hair during the digital dorsal vein image recognition process according to claim 5, characterized in that: In step S4.2, the non-hair region perpendicular to the center line of the hair includes an upper non-hair region and a lower non-hair region. The size of the upper and lower non-hair regions is N times the radius of curvature of the hair point.

7. The method for detecting and processing body hair during digital dorsal vein image recognition according to claim 6, characterized in that: In step S4.2, the repair of each hair point using the non-hair region perpendicular to the hair center line is as follows: grayscale values ​​of pixels in the upper and lower non-hair regions are fitted to obtain the pixel values ​​of the hair point. The fitting weight is inversely proportional to the distance of the pixel from the hair center line, and the fitting formula is: In the formula, Y(n) is the pixel restoration value of the hair point at a distance n from the center line of the hair; R is the radius of curvature of the hair point; α is the fitting weight of the non-hair region. β is the fitting weight for the lower non-hair region; m represents the pixel position in the non-hair area; I0(m) represents the gray value of the pixel in the non-hair area above the hair center line; I1(m) represents the gray value of the pixel in the non-hair area below the hair center line.

8. The method for detecting and processing body hair during the digital dorsal vein image recognition process according to claim 1, characterized in that: The image enhancement process for the hair-removed dorsal vein image in step S5 includes first performing small-scale filtering on the hair-removed dorsal vein image, and then performing large-scale filtering.

Images