A palmprint recognition method and device

Abstract

The present invention relates to a palmprint recognition method and apparatus. The method includes obtaining a palm image to be matched; inputting the palm image into a palm key point model trained by deep learning to segment the palm image into at least two regions of a fixed size; and eliminating a first specific region in the at least two regions of the fixed size through an algorithm of morphological erosion and dilation. The palmprint recognition method and apparatus provided in this application include segmenting the palm image into at least two regions of a fixed size, and eliminating the first specific region included in the at least two regions of the fixed size through the algorithm of morphological erosion and dilation, so as to extract corresponding palmprints from the image after segmentation and elimination and make a comparison, thereby solving the difficulty in palmprint recognition of palm images collected without fixing the palm position in the prior art and improving the efficiency of palmprint recognition.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly to a palmprint recognition method and apparatus. Background Art

[0002] In today's society, personal identity authentication is ubiquitous. Traditional methods such as identity documents cannot be used for personal identity authentication on the Internet. However, methods such as passwords have relatively large security risks. In recent years, biometric recognition technologies such as face recognition, fingerprint recognition, and iris recognition have emerged for identity authentication and have been applied in some fields.

[0003] Most current palmprint recognition technologies adopt a contact-based acquisition method. The contact-based acquisition method uses a closed environment to provide a single background, thus avoiding the influence of the external environment on the acquired information. However, the contact-based acquisition method has problems related to public health. At the same time, the information on the sensor surface in contact may be stolen. Finally, the acquisition environment with a single background for the test is relatively large in volume and not convenient to carry.

[0004] Currently, the commonly used non-contact palmprint extraction method is to perform pixel-level segmentation using methods such as deep learning. However, this method requires continuous training of images in various situations, and the time-consuming for collecting images and training in the early stage is relatively long. At the same time, for the presence of freckles, scars, etc. on the palm, no treatment is carried out, so that these freckles, scars, etc. are extracted as texture features, affecting the accuracy of palmprint recognition. Summary of the Invention

[0005] Based on this, it is necessary to propose a new palmprint recognition method and apparatus to improve the accuracy and speed of palmprint recognition in complex scenarios.

[0006] On the one hand, a specific embodiment of the present application provides a palmprint recognition method, the method comprising:

[0007] Obtaining a palm image to be matched;

[0008] Inputting the palm image into a palm key point model trained by deep learning to segment the palm image into at least two regions of a fixed size;

[0009] Eliminating a first specific region in the at least two regions of a fixed size through morphological erosion and dilation algorithms, the first specific region being a hole and / or gully region in the palm image;

[0010] Performing palmprint recognition on the at least two regions of a fixed size after eliminating the first specific region to determine the information of the palm included in the palm image.

[0011] In a possible design, inputting the palm image into a model trained by deep learning and segmenting the palm image into at least two regions of a fixed size specifically includes:

[0012] Calculating the palm image through the palm key point model to obtain key point coordinate information;

[0013] Segmenting the palm image into at least two regions of a fixed size according to the key point coordinate information.

[0014] In a possible design, the method further includes:

[0015] For any pixel in the at least two regions of a fixed size, obtaining the RGB data of the pixel and the deviation and standard deviation of the RGB data, where the RGB data includes data of the R channel, G channel, and B channel;

[0016] Eliminating the data of the template channel that exceeds the deviation range and filling the eliminated data by the median method to obtain the corrected RGB three-channel data, where the target channel is one of the R channel, G channel, and B channel.

[0017] In a possible design, the method further includes:

[0018] Converting the at least two regions of a fixed size from the RGB data in the RGB color space to the YCrCb data in the YCrCb color space;

[0019] Comparing the converted YCrCb data with a preset YCrCb data threshold to obtain the actual region of the palm in the at least two regions of a fixed size.

[0020] In a possible design, the method further includes:

[0021] Determining the contour of the palm in the at least two regions of a fixed size according to the actual region of the palm in the at least two regions of a fixed size and the region outside the actual region of the palm in the at least two regions of a fixed size;

[0022] Obtaining the largest inscribed circle or largest inscribed rectangle of the contour of the palm corresponding to the palm image in the at least two regions of a fixed size;

[0023] Matching the palm print lines included in the largest inscribed circle or largest inscribed rectangle with the palm print lines of the original palm image stored in advance.

[0024] In a possible design, the method further includes:

[0025] Convert the at least two fixed-size regions from RGB data in the RGB color space into HSB data in the HSB color space, where the HSB data includes H component values;

[0026] Obtain the pixel points in the palm image with the obtained contour where the H component of each pixel exceeds the threshold range;

[0027] Replace the H component of the pixel points in the palm image where the H component exceeds the threshold range with the average H value of the palm image with the obtained contour.

[0028] In a second aspect, a specific embodiment of the present application provides a palmprint recognition device, including an acquisition unit and a processing unit;

[0029] The acquisition unit is used to acquire a palm image to be matched;

[0030] The processing unit is used to input the palm image into a palm key point model trained by deep learning, and segment the palm image into at least two fixed-size regions;

[0031] The processing unit is further used to eliminate a first specific region in the at least two fixed-size regions by an algorithm of morphological erosion and dilation, where the first specific region is a hole and / or gully region in the palm image;

[0032] The processing unit is further used to perform palmprint recognition on the at least two fixed-size regions after eliminating the first specific region to determine the information of the palm included in the palm image.

[0033] In a possible design, the processing unit is further used to:

[0034] For any pixel in the at least two fixed-size regions, obtain the RGB data of the pixel and the deviation and standard deviation of the RGB data, where the RGB data includes data of the R channel, G channel, and B channel;

[0035] Eliminate the data of the template channel that exceeds the deviation range, and fill the eliminated data by the median method to obtain corrected RGB data, where the template channel is one of the R channel, G channel, and B channel.

[0036] In a possible design, the processing unit is further used to:

[0037] Convert the at least two fixed-size regions from RGB data in the RGB color space into YCrCb data in the YCrCb color space;

[0038] Compare the converted YCrCb data with a preset YCrCb data threshold to obtain the actual area of the palm in at least two regions of a fixed size.

[0039] In a possible design, the processing unit is further configured to:

[0040] Convert the at least two regions of a fixed size from RGB data in the RGB color space into HSB data in the HSB color space, where the HSB data includes an H component value;

[0041] Obtain the pixel points in the palm image with the obtained contour where the H component of each pixel point exceeds the threshold range;

[0042] Replace the H component of the pixel points in the palm image where the H component exceeds the threshold range with the average H value of the palm image with the obtained contour.

[0043] In a third aspect, a specific embodiment of the present application further provides a palmprint recognition method, the method includes segmenting the obtained palmprint image and recognizing the segmented palmprint image. The segmentation method includes any one of the first aspect.

[0044] In a fourth aspect, a specific embodiment of the present application further provides a computer device, including a memory and a processor, where computer-readable instructions are stored in the memory, and when the computer-readable instructions are executed by the processor, the processor is caused to execute the following steps:

[0045] Obtain a palm image to be matched; input the palm image into a palm key point model trained by deep learning to segment the palm image into at least two regions of a fixed size; eliminate a first specific region in the at least two regions of a fixed size through morphological erosion and dilation algorithms, where the first specific region is a hole and / or gully region in the palm image; perform palmprint recognition on the at least two regions of a fixed size after eliminating the first specific region to determine the information of the palm included in the palm image.

[0046] In a fifth aspect, a specific embodiment of the present application further provides a storage medium storing computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the one or more processors are caused to execute the steps of any one of the first aspect.

[0047] A palmprint recognition method and device, including dividing a palm image into at least two regions of a fixed size, and eliminating a first specific region included therein from the at least two regions of a fixed size through morphological erosion and dilation algorithms, so as to extract corresponding palmprints from the image after segmentation and elimination and make a comparison, which solves the difficulty in palmprint recognition of palm images collected without fixing the palm position in the prior art and improves the efficiency of palmprint recognition. Description of the Drawings

[0048] Figure 1 A palmprint recognition method provided in a specific embodiment of the present application;

[0049] Figure 2 A palm image calibration method provided in a specific embodiment of the present application;

[0050] Figure 3 A palm image contour determination method provided in a specific embodiment of the present application;

[0051] Figure 4 A method for removing interference from a palm image provided in a specific embodiment of the present application;

[0052] Figure 5 A palmprint recognition device provided in a specific implementation of the present application;

[0053] Figure 6 A computer device provided in a specific embodiment of the present application. Detailed Description of the Invention

[0054] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0055] It can be understood that the terms "first", "second", etc. used in the present application can be used herein to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from another element. For example, without departing from the scope of the present application, the first gesture test script can be called the second gesture test script, and similarly, the second gesture test script can be called the second gesture test script.

[0056] Figure 1 A palmprint recognition method provided in a specific embodiment of the present application. As Figure 2 shown, it includes:

[0057] S1. Obtain a palm image to be matched.

[0058] The palm image can be an image obtained by devices such as cameras and video cameras in a non-fixed scenario or devices including camera and video camera functions. The palm image can include any background and does not include fixed points. The palm image also includes the image information of the palm to be recognized, and the image information of the palm includes palm prints.

[0059] S2. Input the palm image into a palm key point model trained by deep learning to segment the palm image into at least two regions of a fixed size.

[0060] Specifically, the palm key point model trained by deep learning calculates the palm image to obtain the coordinate information of the key points included in the palm image. The coordinate information of the key points is the coordinate information of one or some fixed points in the palm image.

[0061] Segment the palm image into at least one region of a fixed size according to the key point coordinate information.

[0062] In one example, the palm key points of the present application include the midpoint of the root of the little finger, the midpoint of the root of the ring finger, the midpoint of the root of the middle finger, the midpoint of the root of the index finger, and the midpoint of the tip of the middle finger. A total of 5 points are used as palm key points for positioning. Through these 5 key points, the palm image is segmented into a ring finger image region, a middle finger image region, an index finger image region, and a palm image region.

[0063] Of course, in the embodiments of the present application, a traditional skin color (HSV) space domain discrimination method and a method of finding the maximum circumscribed rectangle can also be used to segment the palm image into at least one region of a fixed size.

[0064] S3. Eliminate the first specific region in the at least one region of a fixed size through morphological erosion and dilation algorithms.

[0065] In the embodiments of the present application, the first specific region is the hole and / or gully region in the palm image. By eliminating the holes and / or gullies, a palm image with better quality is obtained.

[0066] The dilation algorithm is a process of merging all background points in contact with an object into the object and expanding the boundary outward. The dilation algorithm can be used to fill holes in an object. The erosion algorithm is a process of eliminating boundary points and contracting the boundary inward. The erosion algorithm can be used to eliminate small and meaningless objects.

[0067] The dilation algorithm can be regarded as the dual operation of the erosion algorithm. Its definition is: translate the structuring element B by a to get Ba. If Ba hits X, we record this a point. The set of all a points that satisfy the above conditions is called the result of X dilated by B. Expressed by the formula: D(X) = {a|Ba↑X} = X B.

[0068] In an example of the dilation algorithm, the pixels of the entire image are traversed in sequence, and the eight surrounding pixels of each pixel are analyzed. As long as there are black pixels around the pixel, the color of the pixel is set to black.

[0069] In an example of the erosion algorithm, each pixel in the image is traversed once, and its eight surrounding pixels are checked. If there are white pixels, the point is set to white.

[0070] Eliminating holes and / or gaps in the palm image through morphological erosion and dilation algorithms is prior art, and this application does not further limit it.

[0071] Eliminating holes and / or gaps in the palm image through morphological erosion and dilation algorithms is only an example in the specific embodiments of this application and cannot be used to limit this application. In one example, opening operation, closing operation, etc. can also be used to eliminate the hole and / or gap area.

[0072] S4. Perform palmprint recognition on the at least two fixed-size regions after eliminating the first specific region to determine the information of the palm included in the palm image.

[0073] The method of palmprint recognition is prior art, and this application does not further limit it.

[0074] This application divides the palm image to be recognized into fixed-size regions, thereby avoiding problems caused by recognizing the entire palm image. At the same time, this application also further improves the quality of the image to be recognized by eliminating holes and / or gaps included in at least one fixed-size region.

[0075] It can be known that each pixel point of each image is data, that is, the picture = 2D data. A color picture can be divided into data of three channels: red, green, and blue, and the value of each point in each channel can be processed separately.

[0076] Optionally, on the basis of the above step 3, the embodiment of this application may further include a step of image calibration. It should be noted that the following steps are steps for processing a single pixel, and in the embodiment of this application, each pixel in the at least two fixed-size regions can be processed according to the following steps. Specifically, it includes:

[0077] S401. Obtain the RGB data of the pixel and the deviation and standard deviation of the RGB data, where the RGB data includes data of the R channel, G channel, and B channel.

[0078] S402, eliminating data of a target channel that exceeds a deviation range, and filling the eliminated data with a median method to obtain corrected RGB data, wherein the target channel is one of the R channel, the G channel, and the B channel.

[0079] The 2σ criterion was used to eliminate data that exceeded the stated deviation range.

[0080] Optionally, based on the above step 3 or step 4, the embodiment of the present application may further include step 5. Step 5 specifically includes:

[0081] S501 : Convert the at least one fixed-size region from RGB data in an RGB color space to YCrCb data in a YCrCb color space.

[0082] YCrCb, or YUV, is mainly used to optimize the transmission of color video signals. The YUV color space is a brightness signal Y and chrominance signals U and V, in which brightness and chrominance are separated. In the RGB space, the skin color of the face is greatly affected by brightness, so it is difficult to separate skin color points from non-skin color points. That is to say, after processing in this space, the skin color points are discrete points with many non-skin color points embedded in the middle. In this application, by converting RGB to YCrCb space, the influence of Y (brightness) can be ignored, because this space is little affected by brightness, and skin color will produce good clustering.

[0083] In an example, the formula to convert a pixel's RGB to YCrCb is as follows:

[0084] Y=0.2990R+0.5870G+0.1140B.

[0085] Cb=-0.1687R-0.3313G+0.5000B+128.

[0086] Cr=0.5000R-0.4187G-0.0813B+128.

[0087] S502: Compare the converted YCrCb data with a preset YCrCb data threshold to obtain an actual area of ​​the palm in the at least one area of ​​fixed size.

[0088] The present application compares the data represented by the YCrCb converted from the palm image with a preset YCrCb space threshold range. When the YCrCb value converted from the palm image is within the threshold range, it can be identified as a palm skin color point.

[0089] According to the actual area of ​​the palm in the at least two fixed-sized areas and the area outside the actual area of ​​the palm in the at least two fixed-sized areas, the contour of the palm in the at least two fixed-sized areas is determined. The contour of the obtained area is searched through YCrCb to obtain a more accurate palm area.

[0090] In the embodiment of the present application, the YCrCb space threshold range of the palm image is also set. The YCrCb space threshold range of the palm image can be obtained by YCrCb statistical analysis of multiple palm images. This application will not elaborate on this.

[0091] Optional, in Figure 5 On the basis of the method shown, it also includes using methods such as Hough circle transformation to find the maximum inscribed circle or maximum inscribed rectangle of the palm image whose contour is re-determined by YCrCb, so as to segment the palm print and obtain the palm print image as the comparison matching picture of the palm print.

[0092] Optionally, based on the above step 3 or step 4 or step 5, the embodiment of the present application may further include step 6. Step 6 specifically includes:

[0093] S601 , converting the at least one fixed-size region from RGB data in an RGB color space into HSB data in an HSB color space, wherein the HSB data includes an H component value.

[0094] In the data represented by HSB, H (hues) represents hue, S (saturation) represents saturation, and B (brightness) represents brightness. The higher the values ​​of S and B in HSB mode, the higher the saturation and brightness, and the stronger and brighter the page color is. It has a quick and eye-catching effect on visual stimulation, but it is not easy to watch for a long time. The S values ​​of the above two colors are close, which is a strong state.

[0095] In an example, the RGB to HSV (HSB) conversion of a pixel can be:

[0096]

[0097]

[0098] u=max

[0099] S602, obtaining each pixel point of the palm image whose H component exceeds a threshold range in the contour-obtained palm image.

[0100] In an embodiment of the present application, it further includes determining the threshold range of the H component of the palm image. The threshold range of the H component in this embodiment can be either a fixed value or an interval, or a set of data composed of multiple fixed values or multiple intervals. In one example, when the threshold range of the H component is a set of data composed of multiple fixed values or multiple intervals, the corresponding H component threshold can be determined according to the average value of the H component of the palm image. The H component threshold is one of the multiple fixed values or a numerical interval among the multiple intervals.

[0101] In the present application, to determine whether the H component exceeds the threshold range of the H component, each H component in the palm image is respectively compared with the threshold range of the H component.

[0102] S603, replacing the H component of the pixel points in the palm image whose H component exceeds the threshold range with the average H value of the palm image from which the contour is obtained.

[0103] The H component value of normal skin color is relatively stable, while the H value of abnormal skin color varies greatly. In this method, the H component is set within a certain threshold range. When the H value of a pixel in the palm image exceeds the threshold range, it is determined as abnormal skin color. The H value of the pixel points with abnormal skin color is replaced with the average H value of the image, so that the abnormal skin color is converted to normal skin color. Thus, it effectively avoids the interference of skin spots, scars, etc. existing in the palm on the recognition of the palm image.

[0104] In a specific embodiment of the present application, a palmprint recognition method is further provided. The method includes:

[0105] Segmenting the acquired palmprint image, and the segmentation method includes Figures 1 to 6 at least one of the contents shown, and the present application will not elaborate on this.

[0106] Performing binarization processing on the segmented palmprint image.

[0107] A picture is composed of a matrix of multiple pixel points, and the processing of the image is the operation on this matrix of pixel points. When it is necessary to change the color of a certain pixel point, the position of this pixel point is found in the matrix of pixel points, such as the x-th row and the y-th column. So the position of this pixel point in this matrix of pixel points can be expressed as (x, y). The color of a pixel point is represented by three color variables of red, green, and blue, and the color of this pixel point can be changed by assigning values to these three variables. For example, changing it to red (255, 0, 0) can be expressed as (x, y, (R = 255, G = 0, B = 0)).

[0108] Binarization makes the grayscale value of each pixel in the pixel matrix of the image be 0 (black) or 255 (white), that is, makes the whole image show only black and white effects. In the grayscale image, the range of grayscale values is 0 to 255, and in the binarized image, the range of grayscale values is 0 or 255.

[0109] For example, a black pixel:

[0110] After binarization, R = 0

[0111] After binarization, G = 0

[0112] After binarization, B = 0

[0113] A white pixel:

[0114] After binarization, R = 255

[0115] After binarization, G = 255

[0116] After binarization, B = 255

[0117] In one example, the binarization method:

[0118] Take the threshold value as 127 (equivalent to the median of 0 to 255, (0 + 255) / 2 = 127), and make the pixels with grayscale values less than or equal to 127 become 0 (black), and those with grayscale values greater than 127 become 255 (white).

[0119] Method 2:

[0120] Calculate the average value avg of the grayscale values of all pixels in the pixel matrix

[0121] (Gray value of pixel point 1 +... + Gray value of pixel point n) / n = Average value avg of pixel points

[0122] Then compare each pixel point with avg one by one. The pixel points less than or equal to avg are 0 (black), and those greater than avg are 255 (white).

[0123] Method 3:

[0124] Use the histogram method (also called the bimodal method) to find the binarization threshold. The histogram is an important feature of the image. The histogram method believes that the image is composed of foreground and background. On the grayscale histogram, both the foreground and background form peaks, and the threshold is at the lowest valley between the two peaks. After obtaining the threshold, compare one by one.

[0125] Perform grayscale processing on the binarized image.

[0126] When performing the grayscale processing, each pixel in the pixel matrix should satisfy the following relationship: R = G = B (where R represents the value of the red variable, G represents the value of the green variable, and B represents the value of the blue variable, and these three values are equal. The "=" here does not mean assignment in programming languages but equality in mathematics). This value at this time is called the grayscale value.

[0127] In one example, the method of grayscale processing can be as follows:

[0128] Method 1:

[0129] The grayscaled R = (R before processing + G before processing + B before processing) / 3

[0130] The grayscaled G = (R before processing + G before processing + B before processing) / 3

[0131] The grayscaled B = (R before processing + G before processing + B before processing) / 3

[0132] Method 2:

[0133] The grayscaled R = R before processing * 0.3 + G before processing * 0.59 + B before processing * 0.11

[0134] The grayscaled G = R before processing * 0.3 + G before processing * 0.59 + B before processing * 0.11

[0135] The grayscaled B = R before processing * 0.3 + G before processing * 0.59 + B before processing * 0.11

[0136] Compare the image after the grayscale processing with the palm image stored in advance to complete the recognition of the palm image.

[0137] Figure 5 A palmprint recognition device provided for the specific implementation of this application. The device includes an acquisition unit 51 and a processing unit 52.

[0138] The acquisition unit 51 is used to acquire the palm image to be matched;

[0139] The processing unit 52 is used to input the palm image into the palm key point model trained by deep learning to segment the palm image into at least two regions of a fixed size;

[0140] The processing unit 52 is further used to eliminate the first specific region in the at least two regions of a fixed size through the algorithms of morphological erosion and dilation. The first specific region is the hole and / or gully region in the palm image. The processing unit is further used to perform palmprint recognition on the at least two regions of a fixed size after eliminating the first specific region to determine the information of the palm included in the palm image.

[0141] In a possible design, the processing unit 52 is further used to obtain RGB data of the pixel and the deviation and standard deviation of the RGB data for any pixel in the at least two fixed-size areas, wherein the RGB data includes data of the R channel, the G channel and the B channel; eliminate the data of the template channel that exceeds the deviation range, and fill the eliminated data with the median method to obtain the corrected RGB data, wherein the target channel is one of the R channel, the G channel and the B channel.

[0142] In a possible design, the processing unit 52 is further used to convert the at least two fixed-size areas from RGB data in the RGB color space into YCrCb data in the YCrCb color space; and compare the converted YCrCb data with a preset YCrCb data threshold to obtain the actual area of ​​the palm in the at least two fixed-size areas.

[0143] For any pixel in the at least two fixed-size areas, RGB data of the pixel and the deviation and standard deviation of the RGB data are obtained, wherein the RGB data includes data of the R channel, the G channel and the B channel; data of the target channel that exceeds the deviation range is eliminated, and the eliminated data is filled with the median method to obtain corrected RGB data, wherein the target channel is one of the R channel, the G channel and the B channel.

[0144] In a possible design, the processing unit 52 is also used to convert the at least two fixed-size areas from RGB data in the RGB color space to HSB data in the HSB color space, wherein the HSB data includes an H component value; obtain each pixel point in the palm image of the obtained contour whose H component exceeds a threshold range; and replace the H component of the pixel point in the palm image whose H component exceeds the threshold range with the average H value of the palm image of the obtained contour.

[0145] Of course, the palmprint recognition device of the present application can also be used to perform Figures 1 to 4 Any of the methods described.

[0146] It should be understood that the division of each unit of the above data acquisition device is only a division of logical functions. In actual implementation, all or part of them can be integrated into a physical entity, or physically separated. And these units can all be implemented in the form of software called by processing elements; they can also all be implemented in the form of hardware; or some units can be implemented in the form of software called by processing elements, and some units can be implemented in the form of hardware. For example, the determination unit can be a separately established processing element, or can be integrated in a certain chip of the base station. In addition, it can also be stored in the memory of the base station in the form of a program, and the function of the first control unit can be called and executed by a certain processing element of the base station. The implementation of other units is similar. It should be noted that the receiving unit can communicate with the terminal through a radio frequency device and an antenna. For example, the base station can receive the information sent by the terminal through the antenna, and the received information is sent to the receiving unit after being processed by the radio frequency device. In addition, the units of the communication device can be integrated together in whole or in part, or can be independently implemented. The processing element mentioned here can be an integrated circuit with the ability to process signals. In the implementation process, each step of the above method or each of the above units can be completed by the integrated logic circuit in the processor element or the instruction in the form of software.

[0147] For example, the above units can be one or more integrated circuits configured to implement the above method, such as: one or more Application Specific Integrated Circuits (ASICs), or, one or more digital signal processors (DSPs), or, one or more Field Programmable Gate Arrays (FPGAs), etc. Again, when a certain unit above is implemented in the form of a processing element scheduling program, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processors that can call programs. Again, these units can be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0148] In one embodiment, Figure 6 A computer device provided for a specific embodiment of the present application, the computer device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. The memory is used to store a program for implementing Figures 1 to 4 the method executed by the method embodiment shown. The processor calls the program and executes the operations of the above method embodiment.

[0149] Certainly, the functions performed by the processor of the above computer device are only examples in this embodiment and cannot be used to limit the computer device in the present application. The computer device of the present application can execute Figures 1 to 4 any step in the method shown.

[0150] In one embodiment, a storage medium storing computer-readable instructions is provided. When the computer-readable instructions are executed by one or more processors, the one or more processors are caused to execute Figures 1 to 4 any step shown.

[0151] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0152] The steps of the methods or algorithms described in combination with the embodiments disclosed herein can be implemented by hardware, software modules executed by a processor, or a combination of both. The software modules can be placed in a random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium well-known in the technical field.

[0153] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0154] The above embodiments only represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention should be subject to the appended claims.

Claims

1. A palmprint recognition method, characterized in that, it includes: Obtain a palm image to be matched; Input the palm image into a palm key-point model trained by deep learning to segment the palm image into at least two regions of a fixed size; Eliminate a first specific region in the at least two regions of a fixed size through an algorithm of morphological erosion and dilation, where the first specific region is a hole and / or gulf region in the palm image; Perform palmprint recognition on the at least two regions of a fixed size after eliminating the first specific region to determine the information of the palm included in the palm image; The step of inputting the palm image into a palm key-point model trained by deep learning to segment the palm image into at least two regions of a fixed size specifically includes: Calculate the palm image through the palm key-point model to obtain key-point coordinate information, where the key points include the midpoint of the root of the little finger, the midpoint of the root of the ring finger, the midpoint of the root of the middle finger, the midpoint of the root of the index finger, and the midpoint of the tip of the middle finger; Segment the palm image into at least two regions of a fixed size according to the key-point coordinate information, and the regions of a fixed size include finger image regions and palm image regions, where the finger image regions include a ring finger image region, a middle finger image region, and an index finger image region; The method further includes: Convert the RGB data of the at least two regions of a fixed size into HSB data, and replace abnormal pixel points in the palm image with normal pixel points, where the abnormal pixel points are pixel points whose H component exceeds the threshold range.

2. The method according to claim 1, characterized in that, the method further includes: For any pixel in the at least two regions of a fixed size, obtain the RGB data of the pixel and the deviation and standard deviation of the RGB data, where the RGB data includes data of the R channel, the G channel, and the B channel; Eliminate the data of the target channel that exceeds the deviation range, and use the median method to fill the eliminated data to obtain corrected RGB data, where the target channel is one of the R channel, the G channel, and the B channel.

3. The method according to claim 1, characterized in that, the method further includes: Convert the RGB data of the at least two regions of a fixed size in the RGB color space into YCrCb data in the YCrCb color space; Compare the converted YCrCb data with a preset YCrCb data threshold to obtain the actual region of the palm in the at least two regions of a fixed size.

4. The method according to claim 3, characterized in that, the method further includes: Determine the contour of the palm in the at least two regions of a fixed size according to the actual region of the palm in the at least two regions of a fixed size and the region outside the actual region of the palm in the at least two regions of a fixed size; Obtain the largest inscribed circle or the largest inscribed rectangle of the contour of the palm corresponding to the palm image in the at least two regions of a fixed size. Match the palm print lines included in the largest inscribed circle or the largest inscribed rectangle with the palm print lines of the original palm image stored in advance.

5. The method according to claim 1, wherein, the method further includes: Converting the RGB data of the at least two fixed-size regions from the RGB color space into HSB data in the HSB color space, and the HSB data includes an H component value; Obtain the pixel points in the palm image with the obtained contour whose H component exceeds the threshold range; Replace the H component of the pixel points in the palm image whose H component exceeds the threshold range with the average H value of the palm image with the obtained contour.

6. A palm print recognition device, wherein, comprising: An acquisition unit for acquiring a palm image to be matched; A processing unit for inputting the palm image into a palm key point model trained by deep learning to segment the palm image into at least two fixed-size regions; The processing unit is further configured to eliminate a first specific region in the at least two fixed-size regions through an algorithm of morphological erosion and dilation, and the first specific region is a hole and / or gully region in the palm image; The processing unit is further configured to perform palm print recognition on the at least two fixed-size regions after eliminating the first specific region to determine the information of the palm included in the palm image; The processing unit is further configured to calculate the palm image through the palm key point model to obtain key point coordinate information, where the key points include the midpoint of the base of the little finger, the midpoint of the base of the ring finger, the midpoint of the base of the middle finger, the midpoint of the base of the index finger, and the midpoint of the tip of the middle finger; Segment the palm image into at least two fixed-size regions according to the key point coordinate information, and the fixed-size regions include finger image regions and palm image regions, where the finger image regions include a ring finger image region, a middle finger image region, and an index finger image region; The processing unit is further configured to convert the RGB data of the at least two fixed-size regions into HSB data, and replace the abnormal pixel points in the palm image with normal pixel points, where the abnormal pixel points are pixel points whose H component exceeds the threshold range.

7. The device according to claim 6, wherein, the processing unit is further configured to: For any pixel in the at least two fixed-size regions, obtain the RGB data of the pixel and the deviation and standard deviation of the RGB data, and the RGB data includes data of an R channel, a G channel, and a B channel; Eliminate the data of the target channel that exceeds the deviation range, and fill the eliminated data by the median method to obtain corrected RGB data, and the target channel is one of the R channel, the G channel, and the B channel.

8. The device according to claim 6, wherein, the processing unit is further configured to: Convert the RGB data of the at least two fixed-size regions from the RGB color space into YCrCb data in the YCrCb color space; Compare the converted YCrCb data with a preset YCrCb data threshold to obtain the actual area of the palm in at least two regions of a fixed size.

9. The device according to claim 6, wherein, the processing unit is further configured to: convert the at least two regions of a fixed size from RGB data in the RGB color space into HSB data in the HSB color space respectively, the HSB data including an H component value; obtain the pixel points in the palm image with the obtained contour where the H component exceeds the threshold range; replace the H component of the pixel points in the palm image where the H component exceeds the threshold range with the average H value of the palm image with the obtained contour.

Images