A palmprint ROI extraction and segmentation method based on palm skeleton points

By combining the palm skeleton point method, using neural networks and active shape models to segment and correct palm images, the problem of poor extraction and segmentation performance and accuracy in traditional methods is solved, and more efficient palm print ROI extraction is achieved.

CN115131833BActive Publication Date: 2024-11-05湖南工商大学
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210788919.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-07-06
Publication Date
2024-11-05
Estimated Expiration
2042-07-06

AI Technical Summary

Technical Problem

In the complex background, the traditional palm-line ROI segmentation method cannot effectively extract the precise area of ​​interest due to factors such as light and noise, resulting in poor extraction and segmentation performance, accuracy and adaptability.

Method used

Using a method combining palm skeleton points, analyzing palm images through neural networks to obtain key skeleton points, using active shape model to separate the front and back scenes, perform perspective transformation to correct palm posture, calculate the central moment and scale bar, intercept the ROI area, and perform lighting normalization. .

Benefits of technology

It improves the accuracy and adaptability of palm line ROI extraction, and can correct strabismus palm line to frontal palm line, thereby facilitating the extraction of ROI and enhancing the extraction speed and performance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115131833B_ABST
    Figure CN115131833B_ABST
Patent Text Reader

Abstract

The embodiment of the present invention provides a palmprint ROI extraction and segmentation method combined with palm skeleton points, which belongs to the field of image processing technology and specifically includes: step 1, using a neural network to analyze a palm image to obtain multiple key skeleton points of the palm; step 2, separating the foreground and background of the palm according to all the key skeleton points and the active shape model, and obtaining the boundary between the palm and the background; step 3, normalizing the palm direction, obtaining the palm posture through the key skeleton points and the palm boundary, and using perspective transformation to correct the tilt of the palm; step 4, obtaining the central moment and scale of the palm according to the skeleton points of the palm after tilt correction; step 5, intercepting the ROI area of ​​the palm according to the central moment and the scale; step 6, normalizing the illumination of the ROI area. Through the scheme of the present invention, the speed, accuracy and adaptability of palmprint ROI extraction are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of image processing technology, and in particular to a palmprint ROI extraction and segmentation method combining palm skeleton points. Background Art

[0002] At present, with the rapid development of information technology and network technology, information security has shown unprecedented importance. Biometric technology has been increasingly widely used due to its unique stability, uniqueness and convenience. Palmprint recognition, as a new and effective biometric technology, has attracted extensive attention from researchers at home and abroad due to its simple sampling, rich image information, high user acceptance, difficulty in forgery, and low noise interference. However, in complex backgrounds, due to the influence of factors such as illumination and noise, traditional palmprint ROI segmentation methods cannot effectively extract the region of interest.

[0003] It can be seen that there is an urgent need for a palmprint ROI extraction and segmentation method that combines palm skeleton points with high extraction and segmentation performance, accuracy and adaptability. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a palmprint ROI extraction and segmentation method combined with palm skeleton points, which at least partially solves the problems of poor extraction and segmentation performance, accuracy and adaptability in the prior art.

[0005] The embodiment of the present invention provides a palmprint ROI extraction and segmentation method combined with palm skeleton points, comprising:

[0006] Step 1: Analyze the palm image using a neural network to obtain multiple key skeleton points of the palm;

[0007] Step 2, separating the foreground and background of the palm according to all the key skeleton points and the active shape model, and obtaining the boundary between the palm and the background;

[0008] Step 3: Normalize the direction of the palm, obtain the palm posture through the key skeleton points and the palm boundary, and use perspective transformation to correct the tilt of the palm;

[0009] Step 4, obtaining the central moment and scale of the palm according to the skeleton points of the palm after tilt correction;

[0010] Step 5, intercepting the ROI area of ​​the palm according to the central moment and the scale;

[0011] Step 6: normalize the illumination of the ROI area.

[0012] According to a specific implementation of the embodiment of the present invention, step 1 specifically includes:

[0013] Step 1.1, use the pre-trained neural network of the hands module in the mediapipe toolkit to obtain the 21 skeleton key point information of the palm;

[0014] Step 1.2, digitally reconstruct the palm using skeleton key point information and mark 21 key skeleton points;

[0015] Step 1.3, make the digitally reconstructed palm skeleton consistent with the palm image.

[0016] According to a specific implementation of an embodiment of the present invention, the 21 key skeleton points include:

[0017] Point 0 is the wrist point;

[0018] {1,2,3,4} are the key points on the wrist to thumb skeleton;

[0019] {5,6,7,8} are the key points on the skeleton from wrist to index finger;

[0020] {9,10,11,12} are the key points on the skeleton from wrist to middle finger;

[0021] {13,14,15,16} are the key points on the skeleton from wrist to ring finger;

[0022] {17,18,19,20} are the key points on the skeleton from wrist to little finger;

[0023] Among them, {4,8,12,16,20} are fingertip points, and {5,9,13,17} are finger root points.

[0024] According to a specific implementation of the embodiment of the present invention, step 2 specifically includes:

[0025] Step 2.1, using all the key skeleton points and the shape proportion of the palm, the palm is modeled with the skeleton points as the center to obtain the initial curve L for palm boundary optimization;

[0026] Step 2.2, obtain the texture and color information of the initial curve L and the area around the point, use the active shape model to optimize and find the boundary curve of the palm, and obtain the boundary between the palm and the background.

[0027] According to a specific implementation of the embodiment of the present invention, step 2.2 specifically includes:

[0028] Step 2.2.1, obtain the grayscale information of 21 key skeleton points and save it as a sample vector, recorded as:

[0029] {g0,g1,g2,……,g 20};

[0030] Step 2.2.2, select n points with the same spacing on the initial curve L, and optimize each point in turn;

[0031] Step 2.2.3, normalize the local features of each point:

[0032]

[0033] Step 2.2.4, calculate the mean of the grayscale model:

[0034]

[0035] Step 2.2.5, calculate the covariance matrix of the grayscale model:

[0036]

[0037] Step 2.2.6, calculate the feature points found during the matching process, calculate their Mahalanobis distance, and find the best matching point:

[0038]

[0039] Step 2.2.7, use multi-resolution images to fit the image boundary;

[0040] Step 2.2.8, repeat steps 2.2.1 to 2.2.7 to obtain the optimized boundary curve L' of the palm.

[0041] According to a specific implementation of the embodiment of the present invention, step 3 specifically includes:

[0042] Step 3.1, combining the key skeleton point information of the palm with the boundary information of the palm, and normalizing the direction and size of the palm using rotation and translation according to the key skeleton point information;

[0043] Step 3.2, compare the palm image with the front view template palm to obtain the parameters of the subsequent perspective transformation;

[0044] Step 3.3, convert the palm image into a front view palm image through perspective transformation.

[0045] According to a specific implementation of the embodiment of the present invention, step 4 specifically includes:

[0046] Step 4.1, obtaining skeleton points of the corrected palm area, and calculating the central moment of the palm area through the skeleton points;

[0047] Step 4.2, obtain the distance L1 between the middle finger root point and the ring finger root point of the palm skeleton point as the scale of the palm.

[0048] According to a specific implementation of the embodiment of the present invention, step 5 specifically includes:

[0049] Step 5.1, take the central moment of the palm as the center point of the ROI, and intercept a square ROI area with a side length of 2L1 and the central moment as the center point;

[0050] Step 5.2: Normalize the size of the image in the ROI area.

[0051] According to a specific implementation of the embodiment of the present invention, step 6 specifically includes:

[0052] Step 6.1, convert the RGB image of the ROI area into a grayscale image;

[0053] Step 6.2, calculate the average grayscale of the grayscale image;

[0054] Step 6.3, divide the grayscale image into 16*16 blocks, and calculate the average grayscale of each block to obtain the brightness matrix E1 of the word block;

[0055] Step 6.4, subtract the average grayscale of the original image from the word block to obtain the brightness matrix E2 of the word block;

[0056] Step 6.5, use bicubic interpolation to reconstruct the mean E into the same value as the original Figure 1 The brightness distribution matrix R of the sample size;

[0057] Step 6.6, subtract the corresponding value in the brightness difference of the whole image from each pixel value in the original image, so as to achieve light brightness balance;

[0058] In step 6.7, use gamma correction to adjust the overall brightness of the image to a specified value.

[0059] The palmprint ROI extraction and segmentation scheme combined with palm skeleton points in the embodiment of the present invention includes: step 1, using a neural network to analyze a palm image to obtain multiple key skeleton points of the palm; step 2, separating the foreground and background of the palm according to all the key skeleton points and the active shape model, and obtaining the boundary between the palm and the background; step 3, normalizing the palm direction, obtaining the palm posture through the key skeleton points and the palm boundary, and using perspective transformation to correct the tilt of the palm; step 4, obtaining the central moment and scale of the palm according to the skeleton points of the palm after tilt correction; step 5, intercepting the ROI area of ​​the palm according to the central moment and the scale; step 6, normalizing the illumination of the ROI area.

[0060] The beneficial effect of the embodiment of the present invention is that through the scheme of the present invention, the oblique palm print image can be corrected to the front view palm print image through angle correction, so as to facilitate the palm print ROI extraction. In the palm print ROI extraction process, the skeleton points of the palm are first marked by a neural network, and then the approximate palm area and morphological information are obtained through the palm skeleton points, and the position of the palm ROI is marked according to the stable skeleton point position, thereby improving the palm print ROI extraction speed, accuracy and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0062] Figure 1 A schematic diagram of a flow chart of a palmprint ROI extraction and segmentation method combined with palm skeleton points provided in an embodiment of the present invention;

[0063] Figure 2 A schematic diagram of a specific application process of a palmprint ROI extraction and segmentation method combined with palm skeleton points provided by an embodiment of the present invention;

[0064] Figure 3 A schematic diagram of 21 key points of a palm skeleton marker provided by an embodiment of the present invention;

[0065] Figure 4 A palm boundary schematic diagram obtained by using an active shape model provided in an embodiment of the present invention;

[0066] Figure 5 A schematic diagram of sequentially optimizing n points with the same spacing on an initial curve L provided by an embodiment of the present invention;

[0067] Figure 6 A corrected frontal palm image provided by an embodiment of the present invention;

[0068] Figure 7 A schematic diagram of ROI region capture provided by an embodiment of the present invention;

[0069] Figure 8 This is a diagram of the illumination preprocessing effect of a ROI region provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0070] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0071] The following describes the embodiments of the present invention through specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0072] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present invention, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this device and / or practice this method.

[0073] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention. The drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0074] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects described may be practiced without these specific details.

[0075] The embodiment of the present invention provides a palmprint ROI extraction and segmentation method combined with palm skeleton points, and the method can be applied to the biometric information recognition process in scenarios such as information analysis and network security.

[0076] See also Figure 1 , is a flow chart of a palmprint ROI extraction and segmentation method combined with palm skeleton points provided by an embodiment of the present invention. Figure 1 and Figure 2 As shown, the method mainly comprises the following steps:

[0077] Step 1: Analyze the palm image using a neural network to obtain multiple key skeleton points of the palm;

[0078] Furthermore, the step 1 specifically includes:

[0079] Step 1.1, use the pre-trained neural network of the hands module in the mediapipe toolkit to obtain the 21 skeleton key point information of the palm;

[0080] Step 1.2, digitally reconstruct the palm using skeleton key point information and mark 21 key skeleton points;

[0081] Step 1.3, make the digitally reconstructed palm skeleton consistent with the palm image.

[0082] Optionally, the 21 key skeleton points include:

[0083] Point 0 is the wrist point;

[0084] {1,2,3,4} are the key points on the wrist to thumb skeleton;

[0085] {5,6,7,8} are the key points on the skeleton from wrist to index finger;

[0086] {9,10,11,12} are the key points on the skeleton from wrist to middle finger;

[0087] {13,14,15,16} are the key points on the skeleton from wrist to ring finger;

[0088] {17,18,19,20} are the key points on the skeleton from wrist to little finger;

[0089] Among them, {4,8,12,16,20} are fingertip points, and {5,9,13,17} are finger root points.

[0090] In the specific implementation, we can first use the pre-trained neural network of the hands module in the mediapipe toolkit to obtain the 21 skeleton key point information of the palm, then use the skeleton point information to digitally reconstruct the palm, mark the 21 skeleton key points, and finally match the digitally reconstructed palm skeleton with the original image. Figure 3 As shown, the process of digitally reconstructing the palm is as follows:

[0091] Step 1: Starting from the bottom of the wrist, number the key points in the order of "from bottom to top, from left to right", starting with 0;

[0092] Step 2: Number the key points of the thumb skeleton from bottom to top as {1, 2, 3, 4};

[0093] Step 3: Number the key points of the index finger skeleton from bottom to top as {5, 6, 7, 8};

[0094] Step 4: Number the key points of the middle finger skeleton from bottom to top as {9, 10, 11, 12};

[0095] Step 5: Number the key points of the ring finger skeleton from bottom to top as {13, 14, 15, 16};

[0096] Step 6: Number the key points of the little finger skeleton from bottom to top as {17, 18, 19, 20};

[0097] In summary, {4,8,12,16,20} are fingertip points, and {5,9,13,17} are finger root points.

[0098] Step 2, separating the foreground and background of the palm according to all the key skeleton points and the active shape model, and obtaining the boundary between the palm and the background;

[0099] Based on the above embodiment, step 2 specifically includes:

[0100] Step 2.1, using all the key skeleton points and the shape proportion of the palm, the palm is modeled with the skeleton points as the center to obtain the initial curve L for palm boundary optimization;

[0101] Step 2.2, obtain the texture and color information of the initial curve L and the area around the point, use the active shape model to optimize and find the boundary curve of the palm, and obtain the boundary between the palm and the background.

[0102] Furthermore, the step 2.2 specifically includes:

[0103] Step 2.2.1, obtain the grayscale information of 21 key skeleton points and save it as a sample vector, recorded as:

[0104] {g0,g1,g2,……,g 20};

[0105] Step 2.2.2, select n points with the same spacing on the initial curve L, and optimize each point in turn;

[0106] Step 2.2.3, normalize the local features of each point:

[0107]

[0108] Step 2.2.4, calculate the mean of the grayscale model:

[0109]

[0110] Step 2.2.5, calculate the covariance matrix of the grayscale model:

[0111]

[0112] Step 2.2.6, calculate the feature points found during the matching process, calculate their Mahalanobis distance, and find the best matching point:

[0113]

[0114] Step 2.2.7, use multi-resolution images to fit the image boundary;

[0115] Step 2.2.8, repeat steps 2.2.1 to 2.2.7 to obtain the optimized boundary curve L' of the palm.

[0116] When implementing it, Figure 4 As shown in , the active shape model (ActivateShapeModel, ASM) is used to optimize and find the boundary curve of the palm. Figure 5 As shown, n points with the same spacing are selected on the initial curve L, and each point is optimized in turn. The specific steps are as follows:

[0117] Step 1: For the kth feature point, find the k-1th and k+1th feature points, connect them, and find the normal line passing through point k;

[0118] Step 2: Take m points with a fixed spacing on both sides of point k on the normal line, and store the grayscale information of these 2m+1 points in a vector as the local feature vector of k;

[0119] Step 3: Repeat the above steps for each feature point to obtain n feature vectors.

[0120] Step 3: Normalize the direction of the palm, obtain the palm posture through the key skeleton points and the palm boundary, and use perspective transformation to correct the tilt of the palm;

[0121] Furthermore, the step 3 specifically includes:

[0122] Step 3.1, combining the key skeleton point information of the palm with the boundary information of the palm, and normalizing the direction and size of the palm using rotation and translation according to the key skeleton point information;

[0123] Step 3.2, compare the palm image with the front view template palm to obtain the parameters of the subsequent perspective transformation;

[0124] Step 3.3, convert the palm image into a front view palm image through perspective transformation.

[0125] In specific implementation, the palm skeleton information and the palm boundary information can be combined first, and the direction and size of the palm can be normalized by rotation and translation according to the skeleton point information; then the palm image is compared with the front view template palm to obtain the parameters of the subsequent perspective transformation; finally, the perspective transformation formula is used:

[0126]

[0127] The palm image is converted into a frontal palm image to solve the tilt problem. The corrected frontal palm image is as follows Figure 6 shown.

[0128] Step 4, obtaining the central moment and scale of the palm according to the skeleton points of the palm after tilt correction;

[0129] Furthermore, the step 4 specifically includes:

[0130] Step 4.1, obtaining skeleton points of the corrected palm area, and calculating the central moment of the palm area through the skeleton points;

[0131] Step 4.2, obtain the distance L1 between the middle finger root point and the ring finger root point of the palm skeleton point as the scale of the palm.

[0132] In the specific implementation, we can first obtain the skeleton points of the corrected palm area, and then use the central moment calculation formula

[0133]

[0134] The central moment of the skeleton point is calculated, and then the distance L1 between the root point of the middle finger bone and the root point of the ring finger bone of the palm skeleton point is obtained as the scale of the palm.

[0135] Step 5, intercepting the ROI area of ​​the palm according to the central moment and the scale;

[0136] Based on the above embodiment, step 5 specifically includes:

[0137] Step 5.1, take the central moment of the palm as the center point of the ROI, and intercept a square ROI area with a side length of 2L1 and the central moment as the center point;

[0138] Step 5.2: Normalize the size of the image in the ROI area.

[0139] In specific implementation, the central moment of the palm can be used as the center point of the ROI, and a square ROI area with a side length of 2L1 and the central moment as the center point can be intercepted, and its inscribed circle can be marked; then the image can be corrected to a horizontal position where the line connecting the middle finger root point and the ring finger root point is connected; finally, the ROI area image is normalized. Figure 7 shown.

[0140] Step 6: normalize the illumination of the ROI area.

[0141] Furthermore, the step 6 specifically includes:

[0142] Step 6.1, convert the RGB image of the ROI area into a grayscale image;

[0143] Step 6.2, calculate the average grayscale of the grayscale image;

[0144] Step 6.3, divide the grayscale image into 16*16 blocks, and calculate the average grayscale of each block to obtain the brightness matrix E1 of the word block;

[0145] Step 6.4, subtract the average grayscale of the original image from the word block to obtain the brightness matrix E2 of the word block;

[0146] Step 6.5, use bicubic interpolation to reconstruct the mean E into the same value as the original Figure 1 The brightness distribution matrix R of the sample size;

[0147] Step 6.6, subtract the corresponding value in the brightness difference of the whole image from each pixel value in the original image, so as to achieve light brightness balance;

[0148] In step 6.7, use gamma correction to adjust the overall brightness of the image to a specified value.

[0149] When implementing it, Figure 8 As shown, Figure 8 The figure shows the lighting preprocessing effect diagram. The steps to normalize the ROI lighting are:

[0150] The first step is to convert the RGB image into a grayscale image;

[0151] The second step is to calculate the average grayscale of the grayscale image;

[0152] The third step is to divide the image into 16*16 blocks, and calculate the average grayscale of each block to obtain the brightness matrix E1 of the word block;

[0153] Step 4: Subtract the average grayscale of the original image from the word block to obtain the brightness matrix E2 of the word block;

[0154] Step 5: Use bicubic interpolation to reconstruct the mean E into the same value as the original Figure 1 The brightness distribution matrix R of the sample size;

[0155] Step 6: Subtract the corresponding value in the brightness difference of the whole image from each pixel value in the original image to achieve light brightness balance;

[0156] In the seventh step, use gamma correction to adjust the overall brightness of the image to the specified value.

[0157] The palmprint ROI extraction and segmentation method combined with palm skeleton points provided in this embodiment can correct the oblique palmprint image into a frontal palmprint image through angle correction, thereby facilitating palmprint ROI extraction. In the palmprint ROI extraction process, the skeleton points of the palm are first marked using a neural network, and then the approximate palm area and morphological information is obtained through the palm skeleton points. The position of the palm ROI is marked according to the stable skeleton point position, thereby improving the palmprint ROI extraction performance, accuracy and adaptability.

[0158] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof.

[0159] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A palmprint ROI extraction and segmentation method combining palm skeleton points, characterized in that: include: Step 1: Analyze the palm image using a neural network to obtain multiple key skeleton points of the palm; Step 2, separating the foreground and background of the palm according to all the key skeleton points and the active shape model, and obtaining the boundary between the palm and the background; The step 2 specifically includes: Step 2.1, using all the key skeleton points and the shape proportion of the palm, the palm is modeled with the skeleton points as the center to obtain the initial curve L for palm boundary optimization; Step 2.2, obtaining the texture and color information of the initial curve L and the area around the point, using the active shape model to optimize and find the boundary curve of the palm, and obtaining the boundary between the palm and the background; The step 2.2 specifically includes: Step 2.2.1, obtain the grayscale information of 21 key skeleton points and save it as a sample vector, recorded as: {g0,g1,g2,……,g 20 }; Step 2.2.2, select n points with the same spacing on the initial curve L, and optimize each point in turn; Step 2.2.3, normalize the local features of each point: Step 2.2.4, calculate the mean of the grayscale model: Step 2.2.5, calculate the covariance matrix of the grayscale model: Step 2.2.6, calculate the feature points found during the matching process, calculate their Mahalanobis distance, and find the best matching point: Step 2.2.7, use multi-resolution images to fit the image boundary; Step 2.2.8, repeat steps 2.2.1 to 2.2.7 to obtain the optimized boundary curve L' of the palm; Step 3: Normalize the direction of the palm, obtain the palm posture through the key skeleton points and the palm boundary, and use perspective transformation to correct the tilt of the palm; Step 4, obtaining the central moment and scale of the palm according to the skeleton points of the palm after tilt correction; Step 5, intercepting the ROI area of ​​the palm according to the central moment and the scale; Step 6: normalize the illumination of the ROI area.

2. The method according to claim 1, characterized in that , the step 1 specifically includes: Step 1.1, use the pre-trained neural network of the hands module in the mediapipe toolkit to obtain the 21 skeleton key point information of the palm; Step 1.2, digitally reconstruct the palm using skeleton key point information and mark 21 key skeleton points; Step 1.3, make the digitally reconstructed palm skeleton consistent with the palm image.

3. The method according to claim 2, characterized in that ,The 21 key skeleton points include: Point 0 is the wrist point; {1,2,3,4} are the key points on the wrist to thumb skeleton; {5,6,7,8} are the key points on the skeleton from wrist to index finger; {9,10,11,12} are the key points on the skeleton from wrist to middle finger; {13,14,15,16} are the key points on the skeleton from wrist to ring finger; {17,18,19,20} are the key points on the skeleton from wrist to little finger; Among them, {4,8,12,16,20} are fingertip points, and {5,9,13,17} are finger root points.

4. The method according to claim 3, characterized in that , the step 3 specifically includes: Step 3.1, combining the key skeleton point information of the palm with the boundary information of the palm, and normalizing the direction and size of the palm using rotation and translation according to the key skeleton point information; Step 3.2, compare the palm image with the front view template palm to obtain the parameters of the subsequent perspective transformation; Step 3.3, convert the palm image into a front view palm image through perspective transformation.

5. The method according to claim 4, characterized in that , the step 4 specifically includes: Step 4.1, obtaining skeleton points of the corrected palm area, and calculating the central moment of the palm area through the skeleton points; Step 4.2, obtain the distance L1 between the middle finger root point and the ring finger root point of the palm skeleton point as the scale of the palm.

6. The method according to claim 5, characterized in that , the step 5 specifically includes: Step 5.1, take the central moment of the palm as the center point of the ROI, and intercept a square ROI area with a side length of 2L1 and the central moment as the center point; Step 5.2: Normalize the size of the image in the ROI area.

7. The method according to claim 6, characterized in that , the step 6 specifically includes: Step 6.1, convert the RGB image of the ROI area into a grayscale image; Step 6.2, calculate the average grayscale of the grayscale image; Step 6.3, divide the grayscale image into 16*16 blocks, and calculate the average grayscale of each block to obtain the brightness matrix E1 of the word block; Step 6.4, subtract the average grayscale of the original image from the word block to obtain the brightness matrix E2 of the word block; Step 6.5, using bicubic interpolation method, reconstruct the mean value E into a brightness distribution matrix R of the same size as the original image; Step 6.6, subtract the corresponding value in the brightness difference of the whole image from each pixel value in the original image, so as to achieve light brightness balance; In step 6.7, use gamma correction to adjust the overall brightness of the image to a specified value.