Finger vein recognition method, device, computer readable storage medium and equipment

Abstract

The application discloses a kind of finger vein identification method, device, computer readable storage medium and equipment, belong to the field of biometric identification.It includes: obtaining finger vein image;Calculate the gradient information of finger vein image, according to gradient information, find the finger boundary using greedy strategy;Detect finger joint on finger vein image, and extract the effective area of finger vein according to the finger joint and finger boundary;Through Gabor filter, the effective area is extracted, and the finger vein feature is obtained;Set sliding window on the finger vein image, and the finger vein feature of the finger vein image in sliding window is compared with the corresponding position of the finger vein feature template Similarity, find the finger vein image in the highest similarity sliding window as the maximum matching area;The similarity of maximum matching area and the corresponding position of finger vein feature template is used as matching score.The application can realize stable and efficient finger vein identification for different devices and different collection environments.

Description

Technical Field

[0001] This invention relates to the field of biometric identification, and in particular to a finger vein identification method, apparatus, computer-readable storage medium, and device. Background Technology

[0002] Biometric identification technology is an operation that verifies identity based on the physiological and behavioral characteristics of the human body. Finger veins are an internal physiological feature of the human body, distributed beneath the skin of the fingers. Collecting finger vein information under near-infrared light can meet the requirements of biometric identification.

[0003] Finger vein recognition boasts excellent stability and security because finger veins are located beneath the skin's surface, eliminating concerns about aging or wear. Furthermore, finger vein acquisition utilizes near-infrared light to illuminate the skin, leveraging the absorption of near-infrared light by hemoglobin to capture vein images, providing natural anti-counterfeiting properties and making it difficult to forge. These advantages have led to increasing attention and widespread application value for finger vein recognition technology.

[0004] The general procedure for finger vein recognition is as follows: Figure 1 As shown, after acquiring finger vein images under near-infrared illumination, finger vein image preprocessing, finger vein feature extraction, and feature comparison are performed sequentially.

[0005] Finger vein image preprocessing: The purpose of finger vein image preprocessing is to extract the effective region containing finger vein information (i.e., the finger vein region) and normalize the scale and enhance the texture of this region. This is because finger vein images are often irregular due to interference from various factors such as acquisition equipment, acquisition environment, and differences in human fingers, and there is also a lot of interfering information in addition to the main part of the finger vein. Finger vein image preprocessing can effectively reduce interference and lay the foundation for extracting effective features.

[0006] Finger vein feature extraction: After obtaining stable finger vein regions through finger vein image preprocessing, an effective method to represent finger vein features is needed for recognition. An effective finger vein feature extraction method is an important means to improve finger vein recognition performance. Common finger vein feature extraction methods in existing technologies include shape-based, texture-based, and minutiae-based methods, among others.

[0007] Feature matching: After extracting the finger vein features, feature matching needs to be performed. In finger vein recognition algorithms, feature matching methods are generally divided into two categories: one is distance-based methods, such as directly calculating the Euclidean distance or cosine distance between two features; the other is classification-based methods, which use machine learning methods such as SVM, neural networks, and fuzzy logic to determine whether two features come from the same category.

[0008] Existing finger vein recognition methods have the following problems:

[0009] 1. In the preprocessing of finger vein images, it is necessary to locate the finger boundary when extracting the effective region containing finger vein information. Current finger vein algorithms are mostly based on grayscale or gradient features to find discrete finger boundary points, which makes the found finger boundary easily affected by noise, reducing the effect of finger boundary detection.

[0010] 2. Due to differences in acquisition equipment / finger placement, the acquired finger vein area may be biased. Traditional methods cannot eliminate the influence of finger vein area bias when extracting the effective area of ​​finger veins, resulting in poor stability and effectiveness of the extracted finger vein area.

[0011] 3. During the finger vein acquisition process, the fingers have a high degree of freedom in placement, which also leads to a certain degree of finger rotation. This results in inconsistent finger postures between two acquisitions, which reduces the vein recognition performance when using traditional finger vein feature extraction and feature comparison algorithms. Summary of the Invention

[0012] To address the technical problem that existing finger vein recognition methods are easily affected by noise and finger placement, this invention provides a finger vein recognition method, device, computer-readable storage medium, and equipment. This invention can achieve stable and efficient finger vein recognition for different devices and different acquisition environments.

[0013] The technical solution provided by this invention is as follows:

[0014] In a first aspect, the present invention provides a finger vein recognition method, the method comprising:

[0015] Acquire images of the finger veins to be detected;

[0016] Calculate the gradient information of the finger vein image, and use a greedy strategy to find the finger boundary based on the gradient information;

[0017] Detect finger joints on the finger vein image and extract the effective area of ​​the finger vein based on the finger joints and finger boundaries;

[0018] The effective region is used to extract features to obtain finger vein features;

[0019] A sliding window is set on the finger vein image, and the similarity of the finger vein features of the finger vein image in the sliding window with the corresponding position of the finger vein feature template is compared. The finger vein image in the sliding window with the highest similarity is selected as the maximum matching region.

[0020] The similarity between the finger vein features of the finger vein image within the maximum matching region and the corresponding position of the finger vein feature template is used as the matching score.

[0021] In a second aspect, the present invention provides a finger vein recognition device, the device comprising:

[0022] The finger vein image acquisition module is used to acquire images of the finger veins to be detected.

[0023] The finger boundary localization module is used to calculate the gradient information of the finger vein image and find the finger boundary using a greedy strategy based on the gradient information.

[0024] The effective region extraction module is used to detect finger joints on the finger vein image and extract the effective region of the finger vein based on the finger joints and finger boundaries;

[0025] The feature extraction module is used to extract features from the effective region using a Gabor filter to obtain finger vein features;

[0026] The sliding comparison module is used to set a sliding window on the finger vein image and compare the similarity of the finger vein features of the finger vein image within the sliding window with the corresponding positions of the finger vein feature template, and find the finger vein image within the sliding window with the highest similarity as the maximum matching region.

[0027] The matching score determination module is used to determine the similarity between the finger vein features of the finger vein image in the maximum matching region and the corresponding position of the finger vein feature template as the matching score.

[0028] Thirdly, the present invention provides a computer-readable storage medium for finger vein recognition, including a memory for storing processor-executable instructions, which, when executed by the processor, implement the steps of the finger vein recognition method described in the first aspect.

[0029] Fourthly, the present invention provides a device for finger vein recognition, comprising at least one processor and a memory storing computer-executable instructions, wherein the processor executes the instructions to implement the steps of the finger vein recognition method described in the first aspect.

[0030] The present invention has the following beneficial effects:

[0031] 1. The present invention uses a greedy strategy for finger boundary detection algorithm. Based on the gradient information of vein images, it quickly and accurately finds a set of continuous finger boundaries through a greedy strategy, which can effectively reduce noise interference in boundary detection.

[0032] 2. This invention is based on the dynamic acquisition of the effective area of ​​finger veins by finger joints. It ensures the effectiveness of the algorithm for different acquisition devices and can extract stable effective areas from images acquired by different devices without adjusting too many parameters, thus improving the generalization of the algorithm.

[0033] 3. This invention uses Gabor encoding to encode finger vein features, performing binary encoding based on the response values ​​of multiple Gabor codes to obtain finger vein features; and utilizes sliding matching to find the common region with the highest similarity (i.e., the maximum matching region) and provides a similarity score. This effectively improves the problems of inconsistent imaging areas and poor matching accuracy caused by finger axial rotation. Attached Figure Description

[0034] Figure 1 A schematic diagram of the existing finger vein recognition process;

[0035] Figure 2 A flowchart illustrating one embodiment of the finger vein recognition method of the present invention;

[0036] Figure 3 This is a schematic diagram of the finger boundary detection process of the present invention;

[0037] Figure 4 This is a flowchart of the finger boundary detection method of the present invention;

[0038] Figure 5 This is a schematic diagram of the process of coarse positioning of the finger boundary according to the present invention;

[0039] Figure 6 This is a flowchart illustrating the coarse positioning of the finger boundary according to the present invention;

[0040] Figure 7 This is a schematic diagram of the rotation correction process of the present invention;

[0041] Figure 8 This is a flowchart of the rotation correction method of the present invention;

[0042] Figure 9 This is a schematic diagram illustrating the process of precise coarse positioning of the finger edge according to the present invention;

[0043] Figure 10 This is a flowchart illustrating the precise positioning of the finger boundary according to the present invention;

[0044] Figure 11 This is a schematic diagram illustrating the process of obtaining the effective area of ​​the present invention;

[0045] Figure 12 A flowchart for obtaining the effective area of ​​the present invention;

[0046] Figure 13A flowchart illustrating another embodiment of the finger vein recognition method of the present invention;

[0047] Figure 14 This is a diagram illustrating the effect of the image enhancement method of the present invention;

[0048] Figure 15 This is a schematic diagram of one embodiment of the finger vein recognition device of the present invention;

[0049] Figure 16 Finger boundary positioning module

[0050] Figure 17 This is a schematic diagram of a coarse positioning unit;

[0051] Figure 18 For rotation correction unit;

[0052] Figure 19 This is a schematic diagram of the effective region extraction module;

[0053] Figure 20 A schematic diagram of a precision positioning unit;

[0054] Figure 21 This is a schematic diagram of another embodiment of the finger vein recognition device of the present invention. Detailed Implementation

[0055] To make the technical problems, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. The components of the embodiments of this invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention.

[0056] Example 1:

[0057] This invention provides a method for recognizing finger veins, such as... Figure 2 As shown, the method includes:

[0058] S100: Acquire the image of the finger vein to be detected.

[0059] In this step, finger vein images are acquired using a bilateral finger vein infrared acquisition device, and the acquired finger vein images are obtained.

[0060] S200: Calculate the gradient information of the finger vein image, and use a greedy strategy to find the finger boundary based on the gradient information.

[0061] In existing technologies, finger boundary detection is easily affected by noise. To solve this problem, this invention uses a greedy strategy for finger boundary detection, which can quickly and accurately find a set of continuous finger boundaries and effectively reduce noise interference in boundary detection.

[0062] Because the acquired finger vein images have clear edge information, the upper / lower boundaries of the finger can be quickly determined by finding the boundaries from dark to light and from light to dark. These boundaries can be represented by the gradient information of the image; therefore, the gradient information of the finger vein image can be calculated, and based on this gradient information, a greedy strategy can be used to find the finger boundaries.

[0063] S300: Detects finger joints on finger vein images and extracts the effective area of ​​finger veins based on finger joints and finger boundaries.

[0064] Due to differences in acquisition devices and finger placement, existing methods for extracting effective areas cannot eliminate these differences, resulting in poor stability and effectiveness of the extracted finger vein effective areas. To solve this problem, this invention dynamically acquires the effective area of ​​finger veins based on the finger joint, ensuring the effectiveness of the algorithm for different acquisition devices. It can extract stable effective areas from images acquired by different devices without needing to adjust too many parameters, thus improving the generalization of the algorithm.

[0065] The aforementioned S200 to S300 are the preprocessing steps for finger vein images.

[0066] S400: The effective region is extracted using a Gabor filter to obtain finger vein features.

[0067] After obtaining the effective region of the finger vein through preprocessing, a method to effectively represent its features is needed. An effective feature extraction technique is crucial for improving finger vein recognition. Finger vein features are represented by the surface texture of an image; therefore, using appropriate texture features can effectively describe its surface information. Gabor wavelets are very similar to the visual stimulus responses of simple cells in the human visual system, exhibiting excellent characteristics in extracting local spatial and frequency domain information of the target. Therefore, this invention proposes a feature extraction method based on Gabor encoding. The real part of Gabor waves can effectively detect the valley-shaped texture region of finger veins, and by encoding the direction of the maximum Gabor response, the texture of the finger vein can be represented.

[0068] The expression for the real part of Gabor is:

[0069]

[0070] Where x′=xcos(θ)+ysin(θ), y′=-xsin(θ)+ycos(θ); λ is the wavelength; θ is the direction parameter; ψ is the phase offset; γ is the spatial aspect ratio; σ is the standard deviation.

[0071] The specific steps of feature extraction are as follows:

[0072] 1) Constructing a Gabor filter

[0073] Generate K Gabor real-part filters with directional parameters as follows:

[0074] Optional, K=8.

[0075] 2) Generate the encoding map

[0076] The effective region is filtered using K Gabor filters respectively, resulting in encoded maps in K directions:

[0077]

[0078] Where T k (x,y) represents the encoded map in the k-th direction; I(x,y) represents the original map of the effective region; Gabor k This represents the k-th filter, where k = 0, 1, ..., K-1; This represents convolution.

[0079] Next, with T tre For the threshold (optional, T) tre =0), perform binary encoding on the encoded map to obtain Gabor encoded features, i.e., finger vein features B. k (x,y).

[0080]

[0081] S500: Set a sliding window on the finger vein image, and compare the similarity between the finger vein features of the finger vein image in the sliding window and the corresponding position of the finger vein feature template. Find the finger vein image in the sliding window with the highest similarity as the maximum matching region.

[0082] During the acquisition of vein images, issues such as finger axial rotation may occur, leading to inconsistent imaging areas. This significantly impacts the accuracy of finger vein feature matching, and traditional finger vein feature extraction and comparison algorithms degrade vein recognition performance. To address this problem, this invention proposes not only a feature extraction method based on Gabor coding but also a matching method based on a sliding window.

[0083] This invention sets up a sliding window that slides across the finger vein image with a certain step size. The portion of the finger vein image within the sliding window is compared with the portion of a pre-stored finger vein feature template within the sliding window. The optimal match is found using a sliding matching method. The optimal match result can be expressed as:

[0084]

[0085] Among them, [x i1 ,x i2 ,y i1 ,y i2 ] represents the common area between the finger vein image and the finger vein feature template (i.e., the portion of the finger vein image and the finger vein feature template located within the sliding window), (x i1 ,y i1 (x) represents the coordinates of the top-left corner of the public area. i2 ,y i2 () represents the coordinates of the bottom right corner of the public area; This represents the maximum matching region, which is the region where the finger vein image and the portion of the finger vein feature template located within the sliding window are most similar to the finger vein features (i.e., Gabor-coded features). The Gabor-coded features represent the portion of the finger vein image and the finger vein feature template located within the sliding window; i = 1, 2.

[0086] S represents the similarity metric function, which can be expressed using the Hamming distance:

[0087]

[0088] in represents XOR; N represents the total number of pixels involved in the matching.

[0089] The similarity score of the maximum matching region is The corresponding score value is calculated based on Gabor coding features.

[0090] S600: The similarity between the finger vein features of the finger vein image in the maximum matching region and the corresponding position of the finger vein feature template is used as the matching score. Based on the matching score, finger vein recognition can be achieved.

[0091] This invention utilizes Gabor coding features to find the maximum matching region. After finding the maximum matching region, the similarity score of the maximum matching region calculated based on Gabor coding features (i.e., the aforementioned...) can be directly applied. The matching score is the similarity between the largest matching region and the corresponding finger vein feature template calculated using Gabor-encoded features.

[0092] As an alternative strategy, after finding the maximum matching region using Gabor encoding, the similarity score of the maximum matching region calculated based on Gabor encoding features is not directly used. Instead of using the matching score, features (such as competing Gabor, HoG features, etc.) are extracted from the maximum matching region, and the feature similarity of the maximum matching region is calculated as the matching score based on the extracted features.

[0093] The present invention has the following beneficial effects:

[0094] 1. The present invention uses a greedy strategy for finger boundary detection algorithm. Based on the gradient information of vein images, it quickly and accurately finds a set of continuous finger boundaries through a greedy strategy, which can effectively reduce noise interference in boundary detection.

[0095] 2. This invention is based on the dynamic acquisition of the effective area of ​​finger veins by finger joints. It ensures the effectiveness of the algorithm for different acquisition devices and can extract stable effective areas from images acquired by different devices without adjusting too many parameters, thus improving the generalization of the algorithm.

[0096] 3. This invention uses Gabor encoding to encode finger vein features, performing binary encoding based on the response values ​​of multiple Gabor codes to obtain finger vein features; and utilizes sliding matching to find the common region with the highest similarity (i.e., the maximum matching region) and provides a similarity score. This effectively improves the problems of inconsistent imaging areas and poor matching accuracy caused by finger axial rotation.

[0097] S200 of the present invention is used for finger boundary detection. Finger boundary detection is a crucial step in the preprocessing process, which is directly related to the quality of the preprocessing and the subsequent finger vein recognition process. Figure 3 A schematic diagram of finger boundary detection is given. Figure 4 A flowchart for finger boundary detection is provided.

[0098] Specifically, S200 includes:

[0099] S210: The finger vein image is scaled down to obtain a scaled image. The gradient information of the scaled image is calculated. Based on the gradient information of the scaled image, a greedy strategy is used to find the coarse positioning boundary of the finger.

[0100] This step is... Figure 3 The finger boundary is roughly positioned as shown.

[0101] S220: Rotate the finger vein image according to the coarse positioning boundary of the finger.

[0102] This step is... Figure 3 The image shown is rotated and corrected.

[0103] S230: Based on the gradient information of the rotated corrected finger vein image, a greedy strategy is used to find the finger boundary.

[0104] This step is... Figure 3 The finger boundary is precisely located as shown.

[0105] This invention uses a boundary-based method to determine the finger contour. First, the finger boundaries are coarsely located on a scaled-down image; then, the original finger vein image is rotated and corrected based on the coarsely located finger boundaries. Finally, the finger boundaries are finely located on the original finger vein image.

[0106] S210 is the coarse positioning process for the finger boundary, and its schematic diagram is shown below. Figure 5 As shown, the flowchart is as follows Figure 6 As shown.

[0107] S210 includes:

[0108] S211: Downsample the finger vein image according to the scaling factor to obtain the scaled image.

[0109] The scaling factor can be denoted as rescale_size, and the scale of the scaled image obtained after sampling is [s_img_h, s_img_w]. Optionally, bilinear interpolation can be used for downsampling.

[0110] S212: Project the finger vein image horizontally and select the row with the largest projection value as the dividing line.

[0111] Since fingers are approximately horizontal, the original digital vein image is projected horizontally. Optionally, the accumulated projection results can be smoothed after horizontal projection, and then the row with the largest projection value can be selected. f This can serve as a general center dividing line, and it can also indicate the position of the fingers to some extent, such as... Figure 5 As shown.

[0112] To reduce noise interference, image smoothing can be performed on the scaled image. Optionally, a 5×5 median filter and a 3×3 mean filter can be used to smooth the scaled image.

[0113] S213: Calculate the gradient information of the scaling image, and divide the gradient information of the scaling image into an upper half and a lower half by the dividing line.

[0114] After image downsampling (or after image smoothing if applicable), the gradient map of the scaled image is calculated to obtain gradient information.

[0115] Optionally, the gradient can be calculated using the vertical Sobel operator or Canny operator (preferably only the vertical gradient). The two lines where the gradient changes abruptly are the finger boundaries, such as... Figure 5 As shown, the brighter line is the upper boundary (grayscale value from low to high, gradient value greater than 0), and the darker line is the lower boundary (grayscale value from high to low, gradient value less than 0).

[0116] Since the gradient value at the upper boundary is greater than 0 while the gradient value at the lower boundary is less than 0, the upper and lower boundaries need to be processed separately. The gradient information is divided into an upper half and a lower half by the boundary line obtained above, and the upper half and the lower half are processed separately.

[0117] S214: For the upper half of the gradient information of the scaled image, only retain the pixels whose vertical gradient is greater than the first filtering threshold, and set the vertical gradient greater than the first truncation threshold as the first truncation threshold to obtain the set of coarsely located pixels at the upper boundary.

[0118] This step filters the valid gradients in the upper half of the gradient information of the scaled image using a dual-threshold strategy. First, the first filtering threshold, gra_tre, is determined. u_1 and the first truncation threshold gra_tre u_2 Then only keep gradients greater than gra_tre u_1 For pixels where the gradient value is greater than gra_tre u_2 Then set its gradient value to gra_tre u_2 .

[0119] You can choose the average gradient value as the first filtering threshold gra_tre u_1 The linear combination of the maximum gradient value and the average gradient value is selected as the first cutoff threshold gra_tre u_2 The specific expression is as follows:

[0120] gra_tre u_1 =mean(g(x,y),g(x,y)>0)

[0121]

[0122] p is a linear combination factor (optional p = 0.5); g(x,y) represents the Sobel gradient information in the vertical direction.

[0123] S215: For the lower half of the gradient information in the scaled image, only retain the vertical gradients that are less than the second filtering threshold gra_tre. d_1 The pixels that are less than the second truncation threshold gra_tre d_2The vertical gradient is set as the second truncation threshold to obtain the set of coarse localization pixels at the lower boundary.

[0124] This step filters the valid gradients in the lower half of the gradient information of the scaled image. Similar to S214, it also uses a double-threshold strategy, retaining only gradients with gradients less than gra_tre. d_1 For pixels whose gradient value is less than gra_tre d_2 Then set its gradient value to gra_tre d_2 .

[0125] Similar to S214, gra_tre d_1 =mean(g(x,y),g(x,y)<0)

[0126] gra_tre d_2 =p·mean(g(x,y),g(x,y)<0)+(1-p)·max(g(x,y),g(x,y)<0)

[0127] In addition to filtering effective gradients using S214 and S215, effective pixels can also be filtered based on the fact that the finger direction is approximately horizontal. This requires that the gradient value in the vertical direction be greater than k times that in the horizontal direction (k = 2 is chosen in this invention) to reduce noise interference.

[0128] S216: In the upper boundary coarse positioning pixel set and the lower boundary coarse positioning pixel set, respectively, a greedy strategy is used to find the line segment with the largest cumulative gradient from one direction to another, so as to obtain the upper boundary and lower boundary of coarse positioning, which are used as the coarse positioning boundary of the finger.

[0129] After obtaining and filtering the gradient information, a greedy strategy can be used to find a continuous curve at the upper and lower boundaries, respectively, as the coarse upper and lower boundaries for finger positioning. Since the finger edge is a continuous line segment, a greedy strategy can be used to find the line segment with the largest continuous gradient magnitude in the gradient information from one direction to another as the finger boundary, thereby effectively reducing noise interference. The direction from one direction to another can be from left to right or from right to left; this invention does not limit this.

[0130] Specifically, taking the coarse positioning of the upper boundary as an example:

[0131] For gradient information g(x,y), based on the estimated finger position (i.e., the boundary line) row f Extract the gradient information including the upper boundary (i.e., the upper half of the gradient map) g. u (x,y)=g[:,0:row f +y δ], where y δ Let g be the size adjustment factor. Let the size of the upper part of the gradient information be (g). u _w,g u _h).

[0132] Initialize the gradient accumulation matrix M = zeros(g u _h,g u _w), the effective boundary quantity matrix O = zeros(g u _h,g u _w), boundary record matrix R = zeros(g u _h,g u _w). And let M(0,:)=g u (0,∶)

[0133] The M, O, and R are updated column by column using the following procedure.

[0134]

[0135] Calculate M(g) u _w-1,:)*O(g u The maximum index of _w-1,:) is used as the y-coordinate of the last boundary point.

[0136] Using the boundary record matrix R, trace all boundary points backward from the last boundary point to obtain the upper boundary point set. This is the coarse upper boundary.

[0137] In the same way, the lower half of the gradient information is extracted, and the gradient value is inverted to trace the lower boundary point set. This is the coarse lower boundary.

[0138] The coarse positioning finger boundary contour is obtained based on the interface between the coarse positioning upper boundary and the coarse positioning lower boundary.

[0139] Due to factors such as finger displacement, the acquired finger vein image may have a certain positional offset. After performing coarse localization edge detection on the scaled image, the original finger vein image can be rotated in a plane based on the upper and lower boundaries of the coarse localization to reduce the inconsistency in the position of the finger vein image.

[0140] Specifically, such as Figure 7 , 8 As shown, S220 includes:

[0141] S221: Determine the set of center points based on the upper and lower boundaries of the coarse positioning.

[0142] This step is used to determine a set of center points based on the coarse upper and lower boundary points of each column. The coordinates of each center point are:

[0143]

[0144] in Represents the x and y coordinates of the center point set.

[0145] S222: Obtain the central line by fitting the set of center points, and calculate the angle θ between the central line and the horizontal direction.

[0146] The least squares method can be used to fit the set of center points to obtain a central line, thereby obtaining the angle θ between the central line and the horizontal direction.

[0147] S223: Rotate the finger vein image counterclockwise by an angle θ so that the center line coincides with the horizontal direction, and restore the finger coarse positioning boundary of the zoomed image to the finger vein image.

[0148] This step involves rotating and performing an affine transformation on the finger vein image: Centered at (0,0), the original finger vein image is rotated counterclockwise by θ until the central line coincides with the horizontal direction. Using the affine transformation and considering the image scaling factor, the coordinates of the upper / lower boundary pixels of the rotated original finger vein image are coarsely located.

[0149]

[0150] After affine transformation, a set of coarse positioning upper / lower boundary point coordinates (i.e., finger coarse positioning boundaries) are obtained on the rotated image. Then, the finger coarse positioning boundaries are restored to the finger vein image.

[0151] The aforementioned coarse positioning of the upper / lower boundary points is not precise, and their coordinates need to be further precisely located.

[0152] like Figure 9 , 10 As shown, S230 includes:

[0153] S231: On the rotated corrected finger vein image, extract the upper sub-image containing the coarse upper boundary and the lower sub-image containing the coarse lower boundary.

[0154] Taking the above boundary as an example, based on the result of the affine transformation, the range of the ordinates of the sub-graph containing the coarsely located upper boundary can be obtained: Based on this, the upper sub-image containing the coarse upper boundary is extracted.

[0155] Optionally, in the cropped sub-image containing the coarse positioning upper boundary, the coarse positioning upper boundary needs to be smoothed. Optionally, smoothing can be performed using a 7×7 median filter and a 7×3 mean filter.

[0156] S232: Calculate the gradient information of the upper sub-image and the lower sub-image respectively.

[0157] This step uses a vertical gradient operator to calculate the gradient map and obtain gradient information. Optionally, this invention uses the following self-designed gradient operator:

[0158]

[0159] S233: For the gradient information of the upper sub-image, only retain the pixels whose vertical gradient is greater than the third screening threshold, and set the vertical gradient greater than the third truncation threshold as the third truncation threshold to obtain the set of pixels with fine positioning of the upper boundary.

[0160] S234: For the gradient information of the lower sub-image, only retain the pixels whose vertical gradient is less than the fourth screening threshold, and set the vertical gradient less than the fourth truncation threshold as the fourth truncation threshold to obtain the set of pixels with fine positioning of the lower boundary.

[0161] S233 to S234 is the gradient screening process, which is similar to the screening method in the coarse positioning of the finger boundary. For details, please refer to the descriptions in S214 and S215.

[0162] S235: Using a greedy strategy, find the line segment with the largest cumulative gradient from one direction to another in the set of finely located pixels at the upper boundary and the set of finely located pixels at the lower boundary, respectively, to obtain the upper and lower boundaries of the finger, which are used as the boundaries of the finger.

[0163] S235 describes the process of precise finger boundary localization using a greedy strategy, which is similar to the method of coarse finger boundary localization. For details, please refer to the description in S216.

[0164] Let the upper boundary pixel obtained by precise positioning be... The lower boundary pixel is Where i = 1, 2, 3, ..., m, and m is the number of valid boundary points detected.

[0165] To ensure stable and effective cropping of the region and to guarantee the stability of the algorithm when acquiring images on different devices, such as... Figure 11 , 12 As shown, S300 of the present invention includes:

[0166] S310: Determine the incisional region of the finger based on the upper and lower boundaries of the finger.

[0167] Based on the precisely located finger vein boundary, the lower bound of the upper boundary's ordinate is selected. and the upper bound of the lower boundary ordinate The internal tangent region of the digital vein area is obtained.

[0168] S320: Detect finger joints in the incisional region of the finger and determine the horizontal range of the effective area based on the number and location of the finger joints; wherein the horizontal range needs to include all finger joints.

[0169] This step involves using a sliding window to determine the finger joints within the finger's incisional region. This is because the joint area has a higher grayscale value, and finding the peak of the accumulated grayscale curve within the window allows for rapid determination of the joint's location. Due to differences in acquisition devices and finger placement, finger vein images may contain information about one or two finger joints.

[0170] If only one peak is detected, the peak is identified as the knuckle position. remember The vertical width of the finger corresponding to the position is

[0171] If two or more peaks are detected, the two peaks with larger peak values ​​and greater distance between them are selected as the two finger joint positions, from left to right, representing the first finger joint. Second finger joint And record express The vertical width of the finger at position (i = 1, 2). In this invention, the left side is defined as the base of the finger, and the right side is defined as the fingertip.

[0172] The horizontal range of the effective area shall be determined based on the following specific circumstances:

[0173] If only one finger joint is detected, and the finger joint is located near the left side of the finger. The horizontal range is then determined to be... Optional,

[0174] If only one finger joint is detected, and the finger joint is located near the right side of the finger. The horizontal range is then determined to be... Optional,

[0175] If only one finger joint is detected, and the finger joint is located near the middle of the finger. The horizontal range is then determined to be... Optional,

[0176] If two finger joints are detected, the horizontal range is determined to be... Optional, in

[0177] S330: Determine the vertical coordinate of the finger midline based on the upper and lower boundaries of the finger, and determine the vertical range of the effective area based on the vertical coordinate of the finger midline and the vertical width of the finger at the finger joint.

[0178] In this step, the new finger midline y-coordinate is determined by the upper / lower boundaries. mid Next, determine the vertical range based on the vertical coordinate of the finger midline and the vertical width of the finger at the finger joint. Optional,

[0179] S340: The area formed by the horizontal and vertical ranges of the effective area is the effective area.

[0180] This step is based on and This allows us to determine the effective range of the intercepted area.

[0181] This invention employs a dynamic effective region size determination strategy, using different strategies based on the differences in finger joints to ensure that a stable effective region is captured and to guarantee the stability of the algorithm when acquiring images on different devices.

[0182] After capturing the effective region of the finger veins, further noise reduction and enhancement of the finger vein texture are required. Traditional finger vein enhancement techniques, such as Gabor enhancement and histogram normalization, are easily affected by strong boundary information during the enhancement process.

[0183] To address the problems existing in current image enhancement technologies, such as Figure 13 As shown, the present invention further includes, prior to S400:

[0184] S300': Image enhancement is performed on the effective region using the following formula:

[0185]

[0186] in, The image is enhanced, and k represents the scale range of the enhanced image.

[0187]

[0188] I(x,y) is the image before enhancement, I(x,y) = L(x,y) × R(x,y), where L(x,y) is the ambient light component in I(x,y), and R(x,y) is the target object reflection component in I(x,y). I(x,y) is estimated by L(x,y) = I(x,y) × F(x,y), and R(x,y) is estimated by I(x,y) = L(x,y) × R(x,y).

[0189] The image before enhancement is denoted as , where is the ambient light component and is the target object reflection component.

[0190] d is a set coefficient.

[0191] The derivation of the above formula is as follows:

[0192] According to Retinex theory, the perceived brightness of an object depends on the ambient lighting and the reflection of the incident light from the object's surface. Therefore, the unenhanced image signal I(x,y) can be expressed as:

[0193] I(x,y)=L(x,y)×R(x,y)

[0194] Where L(x,y) represents the ambient light component; R(x,y) represents the target object reflection component.

[0195] Since R(x,y) carries more detailed information about the image, the formula is equivalent to:

[0196] log(R(x,y))=log(I(x,y))-log(L(x,y))

[0197] L(x,y) can be obtained by Gaussian smoothing the image I(x,y), that is:

[0198] L(x,y)=I(x,y)×F(x,y)

[0199]

[0200] After estimating I(x,y), R(x,y) can be further estimated according to the formula I(x,y)=L(x,y)×R(x,y). To improve contrast and retain richer texture information, linear contrast stretching is performed on the estimated R(x,y).

[0201]

[0202] then It is the enhanced image, where k is the scale range of the enhanced image.

[0203] This invention applies linear contrast enhancement in logarithmic space to significantly enhance finger vein texture and effectively reduce noise interference, laying the foundation for feature extraction. The image enhancement effect of this invention is as follows: Figure 14 As shown, the left side is the original image, and the right side is the image after enhancement.

[0204] Because different fingers exhibit size variations, and even images of the same finger may show size differences, it may be necessary to normalize the size of the effective region after cropping it to obtain a fixed image size, in order to eliminate the effects of some geometric distortions. This is feasible; since the ratio of the horizontal to the vertical extent of the effective region is inconsistent, the vertical height of the effective region can be used to determine the normalization scale factor.

[0205] Optionally, the effective region can be normalized to [100, uni_col] using bilinear interpolation, where

[0206] Example 2:

[0207] This invention provides a finger vein recognition device, such as... Figure 15 As shown, the device includes:

[0208] The finger vein image acquisition module 100 is used to acquire the finger vein image to be detected.

[0209] The finger boundary localization module 200 is used to calculate the gradient information of the finger vein image and find the finger boundary using a greedy strategy based on the gradient information.

[0210] The effective region extraction module 300 is used to detect finger joints on a finger vein image and extract the effective region of the finger vein based on the finger joints and finger boundaries.

[0211] The feature extraction module 400 is used to extract features from the effective region using a Gabor filter to obtain finger vein features.

[0212] The sliding comparison module 500 is used to set a sliding window on the finger vein image and compare the similarity between the finger vein features of the finger vein image in the sliding window and the corresponding position of the finger vein feature template, and find the finger vein image in the sliding window with the highest similarity as the maximum matching region.

[0213] The matching score determination module 600 is used to take the similarity between the finger vein features of the finger vein image in the maximum matching region and the corresponding position of the finger vein feature template as the matching score.

[0214] The aforementioned finger boundary positioning module 200, as shown Figure 16 As shown, it includes:

[0215] The coarse localization unit 210 is used to scale down the finger vein image to obtain a scaled image, calculate the gradient information of the scaled image, and find the coarse localization boundary of the finger using a greedy strategy based on the gradient information of the scaled image.

[0216] The rotation correction unit 220 is used to perform rotation correction on the finger vein image based on the finger coarse positioning boundary.

[0217] The precise positioning unit 230 is used to find the finger boundary using a greedy strategy based on the gradient information of the rotated corrected finger vein image.

[0218] Coarse positioning unit 210, such as Figure 17 As shown, it includes:

[0219] Scaling unit 211 is used to downsample the finger vein image according to a scaling factor to obtain the scaled image.

[0220] The dividing line determination unit 212 is used to project the finger vein image horizontally and select the row with the largest projection value as the dividing line.

[0221] The first gradient calculation unit 213 is used to calculate the gradient information of the scaling image and divide the gradient information of the scaling image into an upper half and a lower half by the dividing line.

[0222] The first filtering unit 214 is used to retain only pixels whose vertical gradient is greater than the first filtering threshold in the upper half of the gradient information of the scaled image, and set the vertical gradient greater than the first truncation threshold as the first truncation threshold to obtain the upper boundary coarse positioning pixel set.

[0223] The second filtering unit 215 is used to retain only pixels whose vertical gradient is less than the second filtering threshold for the lower half of the gradient information of the scaled image, and set the vertical gradient less than the second truncation threshold as the second truncation threshold to obtain the lower boundary coarse positioning pixel set.

[0224] The coarse positioning boundary determination unit 216 is used to find the line segment with the largest cumulative gradient from one direction to another in the upper boundary coarse positioning pixel set and the lower boundary coarse positioning pixel set respectively using a greedy strategy, so as to obtain the coarse positioning upper boundary and coarse positioning lower boundary, which are used as the finger coarse positioning boundary.

[0225] Rotation correction unit 220 Figure 18 As shown, it includes:

[0226] The center point set determination unit 221 is used to determine the center point set based on the coarse positioning upper boundary and the coarse positioning lower boundary.

[0227] The line fitting unit 222 is used to fit the center line based on the center point set and calculate the angle θ between the center line and the horizontal direction.

[0228] The rotation unit 223 is used to rotate the finger vein image counterclockwise by an angle θ so that the center line coincides with the horizontal direction, and restore the finger coarse positioning boundary of the zoomed image to the finger vein image.

[0229] Precision positioning unit 230, such as Figure 20 As shown, it includes:

[0230] The sub-image cropping unit 231 is used to crop an upper sub-image containing a coarse upper boundary and a lower sub-image containing a coarse lower boundary from the rotated corrected finger vein image.

[0231] The second gradient calculation unit 232 is used to calculate the gradient information of the upper sub-image and the lower sub-image respectively.

[0232] The third filtering unit 233 is used to retain only pixels whose vertical gradient is greater than the third filtering threshold for the gradient information of the upper sub-image, and set the vertical gradient greater than the third truncation threshold as the third truncation threshold to obtain the set of pixels with fine positioning of the upper boundary.

[0233] The fourth filtering unit 234 is used to retain only pixels whose vertical gradient is less than the fourth filtering threshold for the gradient information of the lower sub-image, and set the vertical gradient less than the fourth truncation threshold as the fourth truncation threshold to obtain the set of pixels with fine positioning of the lower boundary.

[0234] The precise positioning boundary determination unit 235 is used to find the line segment with the largest cumulative gradient from one direction to another in the upper boundary precise positioning pixel set and the lower boundary precise positioning pixel set respectively using a greedy strategy, so as to obtain the upper boundary and lower boundary of the finger as the finger boundary.

[0235] Effective region extraction module 300 Figure 19 As shown, it includes:

[0236] The incisional region determination unit 310 is used to determine the incisional region of the finger based on the upper and lower boundaries of the finger.

[0237] The horizontal range determination unit 320 is used to detect finger joints in the incisional region of the finger and determine the horizontal range of the effective area based on the number and position of the finger joints.

[0238] The vertical range determination unit 330 is used to determine the vertical coordinate of the finger midline based on the upper and lower boundaries of the finger, and to determine the vertical range of the effective area based on the vertical coordinate of the finger midline and the vertical width of the finger at the finger joint.

[0239] The effective area determination unit 340 defines the effective area as the region composed of the horizontal and vertical ranges of the effective area.

[0240] like Figure 21 As shown, the device of the present invention further includes:

[0241] Image enhancement module 300' is used to enhance the effective region using the following formula:

[0242]

[0243] in, The image is enhanced, and k represents the scale range of the enhanced image.

[0244]

[0245] I(x,y) is the image before enhancement, I(x,y) = L(x,y) × R(x,y), where L(x,y) is the ambient light component in I(x,y), and R(x,y) is the target object reflection component in I(x,y). I(x,y) is estimated by L(x,y) = I(x,y) × F(x,y), and R(x,y) is estimated by I(x,y) = L(x,y) × R(x,y).

[0246]

[0247] d is a set coefficient.

[0248] The device provided in this embodiment of the invention has the same implementation principle and technical effects as the aforementioned method embodiment 1. For the sake of brevity, any parts not mentioned in this device embodiment can be referred to the corresponding content in the aforementioned method embodiment 1. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the aforementioned device and unit can all be referred to the corresponding processes in the aforementioned method embodiment 1, and will not be repeated here.

[0249] Example 3:

[0250] The method described in Embodiment 1 of this specification can implement business logic through a computer program and record it on a storage medium. This storage medium can be read and executed by a computer, achieving the effects of the solution described in Embodiment 1. Therefore, this invention also provides a computer-readable storage medium for finger vein recognition, including a memory for storing processor-executable instructions. When executed by a processor, the instructions implement the steps of the finger vein recognition method of Embodiment 1.

[0251] The storage medium may include a physical device for storing information, typically digitizing the information and then storing it using electrical, magnetic, or optical methods. The storage medium may include: devices that store information using electrical energy, such as various types of memory, like RAM and ROM; devices that store information using magnetic energy, such as hard disks, floppy disks, magnetic tapes, magnetic core memory, bubble memory, and USB flash drives; and devices that store information using optical methods, such as CDs or DVDs. Of course, there are other readable storage media, such as quantum memories and graphene memories.

[0252] The computer-readable storage medium described above, according to the description of method embodiment 1, may also include other implementations. Specific implementations can be found in the description of the relevant method embodiment 1, and will not be elaborated upon here.

[0253] Example 4:

[0254] The present invention also provides a device for finger vein recognition. The device may be a standalone computer, or it may include an actual operating device that uses one or more of the methods or embodiments described in this specification. The device for finger vein recognition may include at least one processor and a memory storing computer-executable instructions. When the processor executes the instructions, it implements the steps of any one or more of the finger vein recognition methods described in Embodiment 1.

[0255] The device described above may also include other implementation methods according to the description of the method or apparatus embodiments. For specific implementation methods, please refer to the description of the relevant method embodiment 1, which will not be repeated here.

[0256] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the scope of the technology disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. All should be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for recognizing finger veins, characterized in that, The method includes: Acquire images of the finger veins to be detected; Calculate the gradient information of the finger vein image, and use a greedy strategy to find the finger boundary based on the gradient information; Detect finger joints on the finger vein image and extract the effective area of ​​the finger vein based on the finger joints and finger boundaries; The effective region is used to extract features to obtain finger vein features; A sliding window is set on the finger vein image, and the similarity of the finger vein features of the finger vein image in the sliding window with the corresponding position of the finger vein feature template is compared. The finger vein image in the sliding window with the highest similarity is selected as the maximum matching region. The similarity between the finger vein features of the finger vein image within the maximum matching region and the corresponding position of the finger vein feature template is used as the matching score. The calculation of gradient information in the finger vein image, and the finding of the finger boundary using a greedy strategy based on the gradient information, includes: The finger vein image is scaled down to obtain a scaled image. The gradient information of the scaled image is calculated. Based on the gradient information of the scaled image, a greedy strategy is used to find the coarse positioning boundary of the finger. The finger vein image is rotated and corrected based on the coarse positioning boundary of the finger. The finger boundary is found using a greedy strategy based on the gradient information of the rotated finger vein image.

2. The finger vein recognition method according to claim 1, characterized in that, The process involves scaling down the finger vein image to obtain a scaled image, calculating the gradient information of the scaled image, and using a greedy strategy to find the coarse finger localization boundary based on the gradient information of the scaled image. This includes: The finger vein image is downsampled according to a scaling factor to obtain the scaled image; The finger vein image is projected horizontally, and the row with the largest projection value is selected as the dividing line; Calculate the gradient information of the scaled image, and divide the gradient information of the scaled image into an upper half and a lower half by the dividing line; For the upper half of the gradient information of the scaled image, retain the pixels whose vertical gradient is greater than the first filtering threshold, and set the vertical gradient greater than the first truncation threshold as the first truncation threshold to obtain the set of coarsely located pixels at the upper boundary. For the lower half of the gradient information of the zoomed image, retain the pixels whose vertical gradient is less than the second filtering threshold, and set the vertical gradient less than the second truncation threshold as the second truncation threshold to obtain the set of coarsely located pixels at the lower boundary. In the upper and lower bound coarse positioning pixel sets, a greedy strategy is used to find the line segment with the largest cumulative gradient from one direction to another, thus obtaining the upper and lower bounds of coarse positioning, which serve as the coarse positioning boundaries of the finger.

3. The finger vein recognition method according to claim 2, characterized in that, The rotation correction of the finger vein image based on the coarse finger positioning boundary includes: Determine the set of center points based on the upper and lower boundaries of the coarse positioning; The central line is obtained by fitting the set of center points, and the angle θ between the central line and the horizontal direction is calculated. The finger vein image is rotated counterclockwise by an angle θ so that the center line coincides with the horizontal direction, and the coarse finger positioning boundary of the zoomed image is restored onto the finger vein image.

4. The finger vein recognition method according to claim 3, characterized in that, The step of finding the finger boundary using a greedy strategy based on the gradient information of the rotation-corrected finger vein image includes: On the rotated corrected finger vein image, an upper sub-image containing the coarse upper boundary and a lower sub-image containing the coarse lower boundary were respectively extracted. Calculate the gradient information of the upper and lower sub-images respectively; For the gradient information of the upper sub-image, retain the pixels whose vertical gradient is greater than the third screening threshold, and set the vertical gradient greater than the third truncation threshold as the third truncation threshold to obtain the set of pixels with fine positioning of the upper boundary. For the gradient information of the lower sub-image, retain the pixels whose vertical gradient is less than the fourth screening threshold, and set the vertical gradient less than the fourth truncation threshold as the fourth truncation threshold to obtain the set of pixels with fine positioning of the lower boundary. In the upper and lower boundary fine-localization pixel sets, a greedy strategy is used to find the line segment with the largest cumulative gradient from one direction to another, thereby obtaining the upper and lower boundaries of the finger, which are used as the finger boundary.

5. The finger vein recognition method according to claim 4, characterized in that, The step of detecting finger joints on a finger vein image and extracting the effective region of the finger vein based on the detected finger joints and the finger boundary includes: Determine the incisional area of ​​the finger based on its upper and lower boundaries; Finger joints are detected in the incisional region of the finger, and the horizontal range of the effective area is determined based on the number and location of the finger joints; wherein, the horizontal range includes all finger joints; The vertical coordinate of the finger midline is determined based on the upper and lower boundaries of the finger, and the vertical range of the effective area is determined based on the vertical coordinate of the finger midline and the vertical width of the finger at the finger joint. The effective area is the region formed by the horizontal and vertical ranges of the effective area.

6. The finger vein recognition method according to any one of claims 1-5, characterized in that, The step of extracting features from the effective region using a Gabor filter to obtain finger vein features includes, prior to: Image enhancement is performed on the effective region using the following formula: in, The image is enhanced, and k represents the scale range of the enhanced image. I(x,y) is the image before enhancement, I(x,y) = L(x,y) × R(x,y), where L(x,y) is the ambient light component in I(x,y), and R(x,y) is the target object reflection component in I(x,y). I(x,y) is estimated by L(x,y) = I(x,y) × F(x,y), and R(x,y) is estimated by I(x,y) = L(x,y) × R(x,y). d is a set coefficient.

7. A finger vein recognition device, characterized in that, The device includes: The finger vein image acquisition module is used to acquire images of the finger veins to be detected. The finger boundary localization module is used to calculate the gradient information of the finger vein image and find the finger boundary using a greedy strategy based on the gradient information. The effective region extraction module is used to detect finger joints on the finger vein image and extract the effective region of the finger vein based on the finger joints and finger boundaries; The feature extraction module is used to extract features from the effective region using a Gabor filter to obtain finger vein features; The sliding comparison module is used to set a sliding window on the finger vein image and compare the similarity of the finger vein features of the finger vein image within the sliding window with the corresponding positions of the finger vein feature template, and find the finger vein image within the sliding window with the highest similarity as the maximum matching region. The matching score determination module is used to determine the similarity between the finger vein features of the finger vein image in the maximum matching region and the corresponding position of the finger vein feature template as the matching score. The finger boundary positioning module includes: The coarse localization unit is used to scale down the finger vein image to obtain a scaled image, calculate the gradient information of the scaled image, and use a greedy strategy to find the coarse localization boundary of the finger based on the gradient information of the scaled image. A rotation correction unit is used to perform rotation correction on the finger vein image based on the finger coarse positioning boundary; A precise positioning unit is used to find the finger boundary using a greedy strategy based on the gradient information of the rotated corrected finger vein image.

8. A computer-readable storage medium for finger vein recognition, characterized in that, It includes a memory for storing processor-executable instructions, which, when executed by the processor, implement the steps of the finger vein recognition method according to any one of claims 1-6.

9. A device for finger vein recognition, characterized in that, It includes at least one processor and a memory storing computer-executable instructions, wherein the processor executes the instructions to implement the steps of the finger vein recognition method according to any one of claims 1-6.

Images