A fingerprint image recognition method and system based on deep learning

Abstract

The present invention proposes a fingerprint image recognition method and system based on deep learning. The fingerprint image recognition method includes performing two model trainings on the deep learning model of the fingerprint sample data processed by grayscale compensation and the original fingerprint sample data to complete the trained deep learning model; performing grayscale compensation processing on the fingerprint reference image at the fingerprint reference image end to obtain the fingerprint reference image after grayscale compensation processing, and obtaining the fingerprint features; performing grayscale compensation processing on the fingerprint edge pixel block of the fingerprint image to be detected at the image end to form the target fingerprint image to be detected; obtaining the target fingerprint image to be detected through the deep learning model for fingerprint feature recognition, obtaining the fingerprint features of the target fingerprint image to be detected, and comparing the fingerprint features of the target fingerprint image to be detected with the target fingerprint features to obtain the fingerprint image recognition result. The system includes modules corresponding to the steps of the method.

Description

Technical Field

[0001] The present invention relates to the technical field of fingerprint image recognition, and in particular to a fingerprint image recognition method and system based on deep learning. Background Art

[0002] Fingerprint recognition technology is a type of biometric technology that recognizes an individual by comparing fingerprint features with fingerprint information in a database. With the development of science and technology, fingerprint recognition technology has been widely used in security authentication, access control systems and other fields. However, in practical applications, due to the diversity of fingerprint samples and the differences in collection equipment, the collected fingerprint images often have problems such as uneven grayscale of boundary pixel blocks, which brings challenges to the extraction and recognition of fingerprint features.

[0003] In order to solve this problem, some studies have proposed methods for grayscale compensation of fingerprint images. These methods mainly improve the grayscale distribution of boundary pixel blocks by enlarging or reducing the pixel blocks of the fingerprint image. However, these methods often only target a single pixel block processing method and ignore the processing effects of pixel blocks in different situations.

[0004] In addition, for deep learning models, the quality and quantity of training data will directly affect the performance of the model. In existing technologies, training data mainly comes from registered user fingerprint reference images, but the quality and quantity of these images are limited, so the performance of the model is also limited. Summary of the invention

[0005] The present invention provides a fingerprint image recognition method and system based on deep learning, which is used to solve the problem in the prior art that adjacent fingerprint lines may have joint images due to sweat stains, water stains or problems with image acquisition equipment, thereby causing low boundary clarity of fingerprint lines and affecting the accuracy of fingerprint recognition:

[0006] The present invention proposes a fingerprint image recognition method based on deep learning, and the fingerprint image recognition method comprises:

[0007] S1: Perform grayscale compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data in the database, and use the fingerprint sample data after grayscale compensation processing and the original fingerprint sample data deep learning model to perform model training twice to complete the trained deep learning model;

[0008] S2: performing grayscale compensation processing on the fingerprint boundary pixel block of the registered user's fingerprint reference image, obtaining a fingerprint reference image after grayscale compensation processing, and obtaining fingerprint features of the fingerprint reference image as target fingerprint features through the deep learning model;

[0009] S3: when the user's fingerprint image to be detected is obtained, grayscale compensation processing is performed on the fingerprint edge pixel blocks of the fingerprint image to be detected, so as to form a target fingerprint image to be detected;

[0010] S4: Input the target fingerprint image to be detected into the deep learning model, the deep learning model obtains the target fingerprint image to be detected for fingerprint feature recognition, obtains the fingerprint features of the target fingerprint image to be detected, and compares the fingerprint features of the target fingerprint image to be detected with the target fingerprint features to obtain the fingerprint image recognition result.

[0011] Furthermore, the fingerprint sample data in the database is subjected to grayscale compensation processing of the fingerprint boundary pixel blocks, and the deep learning model of the fingerprint sample data after grayscale compensation processing and the original fingerprint sample data is used to perform two model trainings, and the trained deep learning model is completed, including:

[0012] S11: extracting fingerprint sample data from a database, and using the fingerprint sample data to perform initial training on a deep learning model for fingerprint image feature recognition, to obtain an initial deep learning model after the initial training is completed;

[0013] S12: expanding and reducing the pixel blocks for the fingerprint sample data used for initial training, and obtaining the comprehensive grayscale compensation value corresponding to the fingerprint sample data used for initial training by expanding and reducing the pixel blocks of the fingerprint sample data;

[0014] S13: performing grayscale compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data used for initial training by using the comprehensive grayscale compensation value to obtain adjusted sample data;

[0015] S14: Performing secondary training on the deep learning model using the adjusted sample data to obtain a trained deep learning model;

[0016] S15: Verify the deep learning model using the fingerprint verification sample to obtain the accuracy of fingerprint image feature recognition of the deep learning model;

[0017] S16: When the accuracy is lower than a preset accuracy threshold, the deep learning model is optimized until the verified accuracy reaches or exceeds the preset accuracy threshold.

[0018] Further, pixel block enlargement and pixel block reduction are performed on the fingerprint sample data used for initial training, and a comprehensive grayscale compensation value corresponding to the fingerprint sample data used for initial training is obtained by performing pixel block enlargement and pixel block reduction on the fingerprint sample data, including:

[0019] S121: Enlarging the pixel blocks of the fingerprint sample data used for initial training, and obtaining the fingerprint sample data after the pixel blocks are enlarged as the first sample data; wherein the size range of the enlarged pixel blocks is 1.5D-3D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training;

[0020] S122: Obtain a first grayscale compensation value according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the first sample data;

[0021] S123: reducing the pixel blocks of the fingerprint sample data used for initial training, and obtaining the fingerprint sample data after the pixel blocks are reduced as the second sample data; wherein the size range of the reduced pixel blocks is 0.1D-0.5D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training;

[0022] S124: Obtain a second grayscale compensation value according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the second sample data;

[0023] S125: Obtain a comprehensive grayscale compensation value using the first grayscale compensation value and the second grayscale compensation value.

[0024] Further, the fingerprint reference image end grayscale compensation processing is performed on the fingerprint boundary pixel block of the recorded user's fingerprint reference image to obtain the fingerprint reference image after the grayscale compensation processing, including:

[0025] S21: extracting the grayscale value of the pixel block of the recorded user's fingerprint reference image;

[0026] S22: extracting the comprehensive grayscale compensation value;

[0027] S23: Obtaining a first grayscale compensation amount corresponding to the grayscale compensation processing of the fingerprint reference image end through the grayscale value of the pixel block of the fingerprint reference image and the comprehensive grayscale compensation value;

[0028] S24: adjusting the pixel values ​​of the fingerprint boundary pixel blocks of the fingerprint reference image to a first grayscale compensation amount to form a fingerprint reference image after grayscale compensation processing.

[0029] Further, when the user's fingerprint image to be detected is obtained, the fingerprint edge pixel block of the fingerprint image to be detected is subjected to grayscale compensation processing at the image end to be detected to form a target fingerprint image to be detected, including:

[0030] S31: extracting the gray value of the pixel block of the fingerprint image to be detected;

[0031] S32: extracting the comprehensive grayscale compensation value;

[0032] S33: obtaining a second grayscale compensation amount corresponding to the grayscale compensation processing of the image to be detected through the grayscale value of the pixel block of the fingerprint image to be detected and the comprehensive grayscale compensation value;

[0033] S34: adjusting the pixel values ​​of the fingerprint boundary pixel blocks of the fingerprint image to be detected to the second grayscale compensation amount to form a target fingerprint image to be detected.

[0034] The present invention proposes a fingerprint image recognition system based on deep learning, and the fingerprint image recognition system comprises:

[0035] Data processing module: Perform grayscale compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data in the database, and use the fingerprint sample data after grayscale compensation processing and the original fingerprint sample data deep learning model to perform two model trainings to complete the trained deep learning model;

[0036] Image acquisition module: performs grayscale compensation processing on the fingerprint boundary pixel block of the registered user's fingerprint reference image, obtains the fingerprint reference image after grayscale compensation processing, and obtains the fingerprint features of the fingerprint reference image through the deep learning model as the target fingerprint features;

[0037] Image forming module: when the user's fingerprint image to be detected is acquired, the fingerprint edge pixel block of the fingerprint image to be detected is subjected to grayscale compensation processing at the end of the image to be detected to form a target fingerprint image to be detected;

[0038] Result acquisition module: input the target fingerprint image to be detected into the deep learning model, the deep learning model obtains the target fingerprint image to be detected for fingerprint feature recognition, obtains the fingerprint features of the target fingerprint image to be detected, and compares the fingerprint features of the target fingerprint image to be detected with the target fingerprint features to obtain the fingerprint image recognition result.

[0039] Furthermore, the data processing module includes:

[0040] Data extraction module: extracts fingerprint sample data from the database, and uses the fingerprint sample data to perform initial training on a deep learning model used for fingerprint image feature recognition, thereby obtaining an initial deep learning model after the initial training is completed;

[0041] Compensation value acquisition module: for the fingerprint sample data used for initial training, pixel block enlargement and pixel block reduction are performed, and the comprehensive gray compensation value corresponding to the fingerprint sample data used for initial training is obtained by enlarging and reducing the pixel block of the fingerprint sample data;

[0042] Data acquisition module: performing grayscale compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data used for initial training by using the comprehensive grayscale compensation value to obtain adjusted sample data;

[0043] Model acquisition module: using the adjusted sample data to perform secondary training on the deep learning model to obtain a trained deep learning model;

[0044] Accuracy acquisition module: using fingerprint verification samples to verify the deep learning model, and obtaining the accuracy of fingerprint image feature recognition of the deep learning model;

[0045] Threshold judgment module: When the accuracy rate is lower than the preset accuracy rate threshold, the deep learning model is optimized until the verified accuracy rate reaches or exceeds the preset accuracy rate threshold.

[0046] Furthermore, the compensation value obtaining module includes:

[0047] The first sample data module: expands the pixel blocks of the fingerprint sample data used for initial training, and obtains the fingerprint sample data after the pixel blocks are expanded as the first sample data; wherein the size range of the expanded pixel blocks is 1.5D-3D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training;

[0048] A first grayscale compensation value module: obtaining a first grayscale compensation value according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the first sample data;

[0049] The second sample data module: performs pixel block reduction on the fingerprint sample data used for initial training, and obtains the fingerprint sample data after the pixel block reduction as the second sample data; wherein the size range of the pixel block reduction is 0.1D-0.5D, and D represents the size of the pixel block of the fingerprint sample data used for initial training;

[0050] A second grayscale compensation value module: obtaining a second grayscale compensation value according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the second sample data;

[0051] Comprehensive grayscale compensation value module: uses the first grayscale compensation value and the second grayscale compensation value to obtain a comprehensive grayscale compensation value.

[0052] Furthermore, the image acquisition module includes:

[0053] Gray value extraction module: extracts the gray value of the pixel block of the recorded user's fingerprint reference image;

[0054] Compensation value extraction module: extract comprehensive grayscale compensation value;

[0055] Grayscale compensation value acquisition module: acquires a first grayscale compensation amount corresponding to the grayscale compensation processing of the fingerprint reference image end through the grayscale value of the pixel block of the fingerprint reference image and the comprehensive grayscale compensation value;

[0056] The pixel block adjustment module is used to adjust the pixel values ​​of the fingerprint boundary pixel blocks of the fingerprint reference image to the first grayscale compensation amount to form a fingerprint reference image after grayscale compensation processing.

[0057] Furthermore, the image forming module comprises:

[0058] Detection module: extracts the gray value of the pixel block of the fingerprint image to be detected;

[0059] Extraction module: extract comprehensive grayscale compensation value;

[0060] An acquisition module: acquiring a second grayscale compensation amount corresponding to the grayscale compensation processing of the image to be detected through the grayscale value of the pixel block of the fingerprint image to be detected and the comprehensive grayscale compensation value;

[0061] Image forming module: adjusts the pixel value of the fingerprint boundary pixel block of the fingerprint image to be detected to the second grayscale compensation amount to form the target fingerprint image to be detected.

[0062] The beneficial effects of the present invention are as follows: by performing grayscale compensation processing on the fingerprint boundary pixel blocks, the clarity and contrast of the fingerprint image can be enhanced, thereby improving the accuracy of fingerprint feature extraction and recognition by the deep learning model; by combining the fingerprint sample data after grayscale compensation processing and the original fingerprint sample data in the deep learning model training process, the deep learning model can have better adaptability and robustness to fingerprint images in different situations; by performing grayscale compensation processing on the fingerprint image to be detected, the influence of problems such as poor fingerprint image quality and insufficient clarity on the fingerprint recognition result can be reduced, thereby improving the security and reliability of fingerprint recognition; using the deep learning model to extract and recognize fingerprint features can greatly improve the recognition efficiency and speed, making fingerprint recognition more convenient and efficient; this fingerprint recognition scheme can be applicable to various scenarios requiring identity authentication, such as mobile phone unlocking, access control systems, etc., and has broad application prospects and market prospects. At the same time, through the application of deep learning algorithms, the system can effectively identify and remove joint images caused by problems such as sweat, water stains or image acquisition equipment, thereby improving the clarity and quality of fingerprint images; fingerprint images with higher clarity will help improve the accuracy and robustness of the fingerprint recognition system, reduce the misrecognition rate, and increase the success rate of fingerprint recognition; deep learning methods can learn and understand the fingerprint image features under various different situations, thereby improving the system's ability to deal with interference factors such as sweat and water stains, so that fingerprint recognition can maintain high accuracy even in complex environments; by solving the problem of low clarity of fingerprint line boundaries, users will get a more stable, fast and reliable recognition experience when using the fingerprint recognition system, improving the overall user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 This is a step diagram of a fingerprint image recognition method based on deep learning according to the present invention;

[0064] Figure 2 This is a module diagram of a fingerprint image recognition system based on deep learning described in the present invention. DETAILED DESCRIPTION

[0065] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0066] Embodiment 1,

[0067] This embodiment provides a fingerprint image recognition method based on deep learning, and the fingerprint image recognition method includes:

[0068] S1: Perform grayscale compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data in the database, and perform two model trainings on the deep learning model using the fingerprint sample data after grayscale compensation processing and the original fingerprint sample data to complete the trained deep learning model; specifically, perform two model trainings on the deep learning model by performing grayscale compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data in the database in two image pixel situations of pixel block enlargement and pixel block reduction to complete the trained deep learning model;

[0069] S2: performing grayscale compensation processing on the fingerprint boundary pixel block of the registered user's fingerprint reference image, obtaining a fingerprint reference image after grayscale compensation processing, and obtaining fingerprint features of the fingerprint reference image as target fingerprint features through the deep learning model;

[0070] S3: when the user's fingerprint image to be detected is obtained, grayscale compensation processing is performed on the fingerprint edge pixel blocks of the fingerprint image to be detected, so as to form a target fingerprint image to be detected;

[0071] S4: Input the target fingerprint image to be detected into the deep learning model, the deep learning model obtains the target fingerprint image to be detected for fingerprint feature recognition, obtains the fingerprint features of the target fingerprint image to be detected, and compares the fingerprint features of the target fingerprint image to be detected with the target fingerprint features to obtain the fingerprint image recognition result.

[0072] The working principle of the above technical solution is: grayscale compensation is performed on the fingerprint sample data in the database, and the processed data and the original data are used to train the deep learning model twice. Here, two different image pixel conditions (enlargement and reduction) are used in combination with the fingerprint boundary pixel blocks for grayscale compensation to increase the robustness and generalization ability of the model; the fingerprint reference image that has been recorded by the user is subjected to grayscale compensation of the fingerprint boundary pixel blocks to obtain the fingerprint reference image after grayscale compensation, and the fingerprint features are extracted as the target fingerprint features through the trained deep learning model; when the user's fingerprint image to be detected is obtained, the fingerprint edge pixel blocks are subjected to grayscale compensation to form the target fingerprint image to be detected; the target fingerprint image to be detected is input into the deep learning model, and the model extracts the fingerprint features of the target fingerprint image to be detected, and compares it with the target fingerprint features to obtain the recognition result of the fingerprint image.

[0073] The effects of the above technical solution are as follows: by grayscale compensation processing of fingerprint sample data and using two deep learning model trainings, the model can better adapt to changes in different lighting, angles and quality, improve the robustness of fingerprint recognition, and effectively cope with complex conditions in real scenes; use the deep learning model to extract fingerprint features, and obtain fingerprint image recognition results by comparing the fingerprint features of the target fingerprint image to be detected with the target fingerprint features. The deep learning model can learn more abstract and advanced feature representations, so it can improve the accuracy of fingerprint image recognition; through two different image pixel conditions (enlargement and reduction) combined with the fingerprint boundary pixel block for grayscale compensation processing, the generalization ability of the model is increased, so that the model can better adapt to fingerprint images of different sizes and resolutions, and improve the generalization ability of fingerprint recognition; based on the deep learning model, it has strong practicality and can be applied to various fingerprint recognition scenarios, such as access control systems, mobile phone unlocking, etc., to provide users with a convenient and safe fingerprint recognition experience. At the same time, traditional fingerprint recognition methods usually require tedious feature engineering, while deep learning models can automatically learn the features of fingerprint images, thereby reducing the difficulty and complexity of feature engineering; deep learning models can automatically adjust model parameters to optimize the performance and accuracy of the model, avoiding the trouble and uncertainty of manually adjusting parameters; and can quickly perform fingerprint recognition, and can process a large amount of fingerprint image data in a short time, making fingerprint recognition more real-time and efficient; the training and reasoning of deep learning models can be performed on general computer hardware, without the need for special fingerprint recognition equipment, thereby reducing the cost and threshold of fingerprint recognition; deep learning models can be trained and optimized for different user groups to adapt to the needs and characteristics of different user groups, thereby achieving personalized customization. By reducing the false recognition rate and improving the anti-interference ability, the overall security of the fingerprint recognition system is improved and the possibility of illegal access is reduced.

[0074] Embodiment 2,

[0075] In this embodiment, the fingerprint sample data in the database is subjected to grayscale compensation processing of the fingerprint boundary pixel blocks, and the deep learning model of the fingerprint sample data after grayscale compensation processing and the original fingerprint sample data is used to perform two model trainings to complete the trained deep learning model, including:

[0076] S11: extracting fingerprint sample data from a database, and using the fingerprint sample data to initially train a deep learning model for fingerprint image feature recognition, to obtain an initial deep learning model after the initial training; wherein the deep learning model adopts a model with a convolutional neural network structure;

[0077] S12: expanding and reducing the pixel blocks for the fingerprint sample data used for initial training, and obtaining the comprehensive grayscale compensation value corresponding to the fingerprint sample data used for initial training by expanding and reducing the pixel blocks of the fingerprint sample data;

[0078] S13: grayscale compensation processing is performed on the fingerprint boundary pixel block of the fingerprint sample data for initial training by using the comprehensive grayscale compensation value to obtain adjusted sample data; specifically, the comprehensive grayscale compensation value is obtained by using the first grayscale compensation value and the second grayscale compensation value, and the comprehensive data compensation value is used to perform grayscale compensation processing on the fingerprint boundary pixel block of the fingerprint sample data for initial training, so that the grayscale value of the fingerprint boundary pixel block reaches the comprehensive grayscale compensation value;

[0079] S14: Performing secondary training on the deep learning model using the adjusted sample data to obtain a trained deep learning model;

[0080] S15: Verify the deep learning model using the fingerprint verification sample to obtain the accuracy of fingerprint image feature recognition of the deep learning model;

[0081] S16: When the accuracy is lower than a preset accuracy threshold, the deep learning model is optimized until the verified accuracy reaches or exceeds the preset accuracy threshold.

[0082] The working principle of the above technical solution is: extracting fingerprint sample data from a database, and using the fingerprint sample data to perform initial training on a deep learning model used for fingerprint image feature recognition, to obtain an initial deep learning model after completing the initial training; wherein the deep learning model adopts a model with a convolutional neural network structure; performing pixel block enlargement and pixel block reduction on the fingerprint sample data used for initial training, and obtaining a comprehensive grayscale compensation value corresponding to the fingerprint sample data used for initial training by performing pixel block enlargement and reduction on the fingerprint sample data; performing grayscale compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data used for initial training by using the comprehensive grayscale compensation value to obtain adjusted sample data; specifically The first grayscale compensation value and the second grayscale compensation value are used to obtain a comprehensive grayscale compensation value, and the comprehensive data compensation value is used to perform grayscale compensation processing on the fingerprint boundary pixel block of the fingerprint sample data used for initial training, so that the grayscale value of the fingerprint boundary pixel block reaches the comprehensive grayscale compensation value; the deep learning model is trained again using the adjusted sample data to obtain a deep learning model after training; the deep learning model is verified using a fingerprint verification sample to obtain the accuracy of fingerprint image feature recognition of the deep learning model; when the accuracy is lower than a preset accuracy threshold, the deep learning model is optimized until the accuracy after verification reaches or exceeds the preset accuracy threshold.

[0083] The effects of the above technical solution are as follows: by performing grayscale compensation processing on the fingerprint boundary pixel blocks, the grayscale value of the fingerprint boundary pixel blocks can reach the comprehensive grayscale compensation value, so as to better adapt to the characteristics of the fingerprint image during the deep learning model training process, and improve the accuracy of fingerprint image feature recognition; by performing comprehensive grayscale compensation processing on the fingerprint sample data, the robustness of the model to factors such as different lighting conditions and fingerprint quality differences can be enhanced, and the stability and reliability of the fingerprint recognition system in practical applications can be improved; the technical solution includes a monitoring and optimization mechanism for the model verification accuracy. When the verified accuracy is lower than the preset threshold, the deep learning model can be optimized to adapt it to the characteristics of different data sets, thereby improving the applicability and performance of the system in practical applications; through secondary model training, the model can better adapt to the adjusted sample data, improve the generalization ability of the model, and enable it to achieve better recognition effects when processing new fingerprint sample data. By enlarging and reducing the pixel blocks of the fingerprint sample data and using the comprehensive grayscale compensation value to perform grayscale compensation processing on the fingerprint boundary pixel blocks, the clarity and contrast of the fingerprint image can be enhanced, thereby improving the accuracy of the deep learning model in extracting and identifying fingerprint features; by performing grayscale compensation processing on the fingerprint image to be detected, the impact of problems such as poor fingerprint image quality and insufficient clarity on the fingerprint recognition results can be reduced, thereby improving the security and reliability of fingerprint recognition; using a deep learning model to extract and identify fingerprint features can greatly improve the recognition efficiency and speed, making fingerprint recognition more convenient and efficient. When the accuracy of the deep learning model is lower than the preset accuracy threshold, it can be automatically optimized until the accuracy after verification reaches or exceeds the preset accuracy threshold, thereby ensuring the performance and accuracy of the deep learning model.

[0084] Embodiment 3,

[0085] In this embodiment, pixel block enlargement and pixel block reduction are performed on the fingerprint sample data used for initial training, and the comprehensive grayscale compensation value corresponding to the fingerprint sample data used for initial training is obtained by enlarging and reducing the pixel blocks of the fingerprint sample data, including:

[0086] S121: Enlarging the pixel blocks of the fingerprint sample data used for initial training, and obtaining the fingerprint sample data after the pixel blocks are enlarged as the first sample data; wherein the size range of the enlarged pixel blocks is 1.5D-3D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training;

[0087] S122: Obtain a first grayscale compensation value according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the first sample data; wherein the first grayscale compensation value is obtained by the following formula:

[0088] Hxa =ln(1+H x2 -H x1 )

[0089]

[0090] Among them, H xa Represents the first grayscale compensation value; H x1 and H x2 denote the first compensation component and the second compensation component respectively; m denotes the number of pixel blocks after the size of the pixel block is enlarged; n denotes the number of pixel blocks before the size of the pixel block is enlarged; H di H represents the grayscale value of the i-th pixel block after the pixel block size is enlarged; yi H represents the grayscale value of the i-th pixel block before the size of the pixel block is enlarged; b Represents the average gray value of the pixel block in the background part; sqrt represents square root operation;

[0091] S123: reducing the pixel blocks of the fingerprint sample data used for initial training, and obtaining the fingerprint sample data after the pixel blocks are reduced as the second sample data; wherein the size range of the reduced pixel blocks is 0.1D-0.5D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training;

[0092] S124: Obtain a second grayscale compensation value according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the second sample data; wherein the second grayscale compensation value is obtained by the following formula:

[0093] H xb =ln(1+H x3 -H x2 )

[0094]

[0095] Among them, H xb Represents the second grayscale compensation value; H x3 represents the third compensation component; k represents the number of pixel blocks after the size of the pixel block is reduced; H si Represents the grayscale value of the i-th pixel block after the size of the pixel block is reduced;

[0096] S125: Obtain a comprehensive grayscale compensation value using the first grayscale compensation value and the second grayscale compensation value. The comprehensive grayscale compensation value is obtained by the following formula:

[0097]

[0098] Among them, H z Indicates the comprehensive grayscale compensation value; H pRepresents the average grayscale value of the fingerprint part of each fingerprint sample data.

[0099] The working principle of the above technical solution is as follows: for the fingerprint sample data used for initial training, pixel blocks are enlarged to obtain the fingerprint sample data after the pixel blocks are enlarged as the first sample data; wherein the size range of the enlarged pixel blocks is 1.5D-3D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training; a first grayscale compensation value is obtained according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the first sample data; pixel blocks are reduced to obtain the fingerprint sample data after the pixel blocks are reduced as the second sample data; wherein the size range of the reduced pixel blocks is 0.1D-0.5D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training; a second grayscale compensation value is obtained according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the second sample data; The fingerprint sample data for initial training is reduced in pixel blocks to obtain the fingerprint sample data after the pixel blocks are reduced as the second sample data; wherein the size range of the reduced pixel blocks is 0.1D-0.5D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training; the second grayscale compensation value is obtained according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the second sample data; the fingerprint sample data for initial training is reduced in pixel blocks to obtain the fingerprint sample data after the pixel blocks are reduced as the second sample data; wherein the size range of the reduced pixel blocks is 0.1D-0.5D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training; the second grayscale compensation value is obtained according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the second sample data; the comprehensive grayscale compensation value is obtained by using the first grayscale compensation value and the second grayscale compensation value. The working principle of the above technical solution is:

[0100] The effect of the above technical solution is: the comprehensive grayscale compensation value is obtained by changing the pixel block size, which can effectively improve the accuracy of the comprehensive grayscale compensation value. By expanding and reducing the pixel block of the fingerprint sample data, the fingerprint image information at different scales can be obtained, so that the system can more comprehensively capture the characteristics and details of the fingerprint image; by calculating the first grayscale compensation value and the second grayscale compensation value, the grayscale difference between the fingerprint part and the background part at different scales is compensated, so as to more accurately describe the grayscale characteristics of the fingerprint image at different scales; using the first grayscale compensation value and the second grayscale compensation value, the comprehensive grayscale compensation value is calculated, which can comprehensively consider the grayscale characteristics of the fingerprint image at different scales, and improve the adaptability of the fingerprint recognition system to fingerprint images of various scales; by comprehensively utilizing multi-scale information and grayscale compensation values, the fingerprint recognition system has stronger robustness and accuracy when facing factors such as different scales and lighting conditions, and improves the practical application value of the system; by performing multi-scale processing on the initial training fingerprint sample data, more comprehensive and diversified data support can be provided for model training, which helps to improve the generalization ability and accuracy of the fingerprint recognition system. The above-mentioned first grayscale compensation value formula takes into account multi-scale factors and grayscale value differences, comprehensively considers the grayscale characteristics of the fingerprint part and the background part, enhances the processing capability and accuracy of multi-scale fingerprint images, and helps to improve the robustness and accuracy of the fingerprint recognition system. At the same time, the first compensation component and the second compensation component in the formula respectively consider the grayscale value difference before and after the pixel block size is enlarged. By comparing the grayscale values ​​of pixel blocks of different sizes, the grayscale changes of the fingerprint part and the background part are more comprehensively described; the m and n parameters in the formula respectively represent the number of pixel blocks after the pixel block size is enlarged and reduced. By considering the grayscale values ​​of pixel blocks of different scales, the grayscale compensation value can fully integrate the fingerprint image information at different scales, thereby improving the system's processing capability for multi-scale fingerprint images; the H in the formula b The parameter represents the average grayscale value of the pixel block in the background part. By introducing the grayscale value of the background part into the calculation, the grayscale features of the fingerprint part and the background part can be better distinguished, and the accuracy of the grayscale compensation value is enhanced. The sqrt operation in the formula is used to amplify and balance the grayscale value difference, so that the grayscale compensation value can better reflect the grayscale changes at different scales, and improve the sensitivity to the grayscale features of the fingerprint image. The above second grayscale compensation value calculation formula comprehensively considers multi-scale factors and grayscale value differences. By processing the grayscale values ​​of pixel blocks at different scales, it enhances the processing capability and accuracy of multi-scale fingerprint images, which helps to improve the robustness and accuracy of the fingerprint recognition system. At the same time, the third compensation component H in the formula x3The grayscale value difference after the pixel block size is reduced is comprehensively considered. By processing the grayscale value of the reduced-size pixel block, the grayscale change of the fingerprint part and the background part is more comprehensively described. The k parameter in the formula represents the number of pixel blocks after the pixel block size is reduced. By considering the grayscale values ​​of pixel blocks of different scales, the second grayscale compensation value can fully integrate the fingerprint image information at different scales, thereby improving the system's processing ability for multi-scale fingerprint images. The H in the formula si The parameter represents the grayscale value of the ith pixel block after the pixel block size is reduced. By processing the grayscale value of the reduced-size pixel block, the second grayscale compensation value can better reflect the grayscale changes at different scales, thereby enhancing the sensitivity to the grayscale features of the fingerprint image. The above-mentioned comprehensive grayscale compensation value formula comprehensively considers the grayscale features of the fingerprint sample data, globally processes the grayscale information of the fingerprint image, and reduces the impact of individual samples on the system, which helps to improve the fingerprint recognition system's processing capability and accuracy for different fingerprint samples and enhances the robustness of the system. At the same time, H in the formula p The parameter represents the average grayscale value of the fingerprint part of each fingerprint sample data. By introducing the average grayscale value of each fingerprint sample data, the comprehensive grayscale compensation value can take into account the grayscale feature differences between different fingerprint samples, thereby enhancing the processing capability of different fingerprint samples. z By processing the average grayscale value of the fingerprint part of each fingerprint sample data, the grayscale characteristics of the entire fingerprint image are comprehensively considered, so that the comprehensive grayscale compensation value can more comprehensively describe the grayscale changes between the fingerprint part and the background part. By introducing the average grayscale value of each fingerprint sample data, the formula can better process the grayscale characteristics of individual samples, reduce the impact of individual samples on the entire system, and improve the stability and robustness of the system.

[0101] Embodiment 4,

[0102] In this embodiment, the fingerprint reference image end grayscale compensation processing is performed on the fingerprint boundary pixel block of the registered user's fingerprint reference image to obtain the fingerprint reference image after the grayscale compensation processing, including:

[0103] S21: extracting the grayscale value of the pixel block of the recorded user's fingerprint reference image;

[0104] S22: extracting the comprehensive grayscale compensation value;

[0105] S23: Obtaining a first grayscale compensation amount corresponding to the grayscale compensation processing of the fingerprint reference image end through the grayscale value of the pixel block of the fingerprint reference image and the comprehensive grayscale compensation value; wherein the first grayscale compensation amount is obtained through the following formula:

[0106]

[0107]

[0108] Among them, H 01 represents the first grayscale compensation amount; r represents the number of pixel blocks of the fingerprint reference image; Hri represents the grayscale value of the i-th pixel block of the fingerprint reference image; H z represents the comprehensive grayscale compensation value; C1 represents the number of pixel blocks of the fingerprint part of the fingerprint reference image; C2 represents the number of pixel blocks of the background part of the fingerprint reference image; λ1 represents the first adjustment coefficient;

[0109] S24: adjusting the pixel values ​​of the fingerprint boundary pixel blocks of the fingerprint reference image to a first grayscale compensation amount to form a fingerprint reference image after grayscale compensation processing.

[0110] The working principle of the above technical solution is: extract the grayscale value of the pixel block of the fingerprint reference image that has been recorded by the user; the system will process the fingerprint reference image that has been recorded by the user, divide the image into different pixel blocks, and extract the grayscale value of each pixel block. These grayscale values ​​reflect the brightness or grayscale level of each area in the fingerprint image; the extraction of the comprehensive grayscale compensation value may involve analyzing the entire fingerprint reference image to determine the overall grayscale characteristics and compensation requirements. This value may be calculated based on factors such as the overall brightness and contrast of the image, and is used for subsequent grayscale compensation processing. By combining the grayscale value of the pixel block of the fingerprint reference image and the comprehensive grayscale compensation value, the system calculates a first grayscale compensation amount, which is usually used for grayscale adjustment of the entire image to achieve overall grayscale compensation processing. Finally, the system adjusts the pixel value of the fingerprint boundary pixel block of the fingerprint reference image to the first grayscale compensation amount to form a fingerprint reference image after grayscale compensation processing. This step is equivalent to fine grayscale adjustment of the local area of ​​the image to enhance the characteristics of the fingerprint and improve the quality of the image.

[0111] The effects of the above technical solution are: by performing grayscale compensation processing on the fingerprint reference image, the interference caused by external factors such as lighting conditions and influences can be effectively reduced, thereby improving the accuracy and robustness of fingerprint recognition; grayscale compensation processing can help highlight and enhance the features in the fingerprint image, making the fingerprint image more distinguishable, thereby improving the reliability of fingerprint recognition; by adjusting the pixel value of the fingerprint reference image, the grayscale compensation processing helps to optimize the quality of the fingerprint image, making the image clearer, easier to analyze and identify; the extraction of the comprehensive grayscale compensation value and the adjustment of the grayscale compensation amount can be adjusted according to different environmental conditions and needs, so that the fingerprint recognition system can better adapt to different working scenarios. The above-mentioned first grayscale compensation amount calculation formula can make the calculation of the grayscale compensation amount more intelligent and personalized, and can better adapt to the characteristics of different fingerprint images, thereby improving the performance of the entire fingerprint recognition system. At the same time, the above formula can be personalized according to the characteristics of the specific fingerprint reference image. By considering the number of pixel blocks in the fingerprint part and the background part, the formula can compensate the grayscale features of different parts to different degrees to adapt to the characteristics of different fingerprint images. The formula contains the comprehensive grayscale compensation value H z , which enables the calculation of grayscale compensation to comprehensively consider the grayscale features of the entire fingerprint image, rather than just being limited to the grayscale values ​​of the local area. This allows for a more comprehensive adjustment of the image's grayscale, improving the quality of the fingerprint image. The introduction of the adjustment coefficient λ1 makes the grayscale compensation calculation in the formula adjustable, and the parameters can be adjusted according to actual conditions to obtain the best grayscale compensation effect. Through this personalized, comprehensively considered and adjustable grayscale compensation calculation, the grayscale of the fingerprint reference image can be adjusted more accurately, thereby improving the accuracy and robustness of the fingerprint recognition system for fingerprint images.

[0112] Embodiment 5,

[0113] In this embodiment, when a fingerprint image to be detected of a user is obtained, a grayscale compensation process is performed on a fingerprint edge pixel block of the fingerprint image to be detected at the image end to be detected to form a target fingerprint image to be detected, including:

[0114] S31: extracting the gray value of the pixel block of the fingerprint image to be detected;

[0115] S32: extracting the comprehensive grayscale compensation value;

[0116] S33: Obtaining a second grayscale compensation amount corresponding to the grayscale compensation processing of the image to be detected through the grayscale value of the pixel block of the fingerprint image to be detected and the comprehensive grayscale compensation value; wherein the second grayscale compensation amount is obtained by the following formula:

[0117]

[0118]

[0119] Among them, H 02 represents the second grayscale compensation amount; t represents the number of pixel blocks of the fingerprint image to be detected; Hti represents the grayscale value of the i-th pixel block of the fingerprint image to be detected; H z represents the comprehensive grayscale compensation value; C3 represents the number of pixel blocks of the fingerprint part of the fingerprint image to be detected; C4 represents the number of pixel blocks of the background part of the fingerprint image to be detected; λ2 represents the second adjustment coefficient.

[0120] S34: adjusting the pixel values ​​of the fingerprint boundary pixel blocks of the fingerprint image to be detected to the second grayscale compensation amount to form a target fingerprint image to be detected.

[0121] The working principle of the above technical solution is: the fingerprint image to be detected is segmented into pixel blocks, each pixel block contains a certain number of pixels. Then, the grayscale values ​​of these pixel blocks are extracted. The grayscale value represents the color information of each pixel, ranging from 0 to 255, where 0 represents black and 255 represents white. This process is to obtain the brightness information of each pixel in the image; the extraction of the comprehensive grayscale compensation value is to consider the brightness and contrast of the entire image. It is obtained by processing various statistical features such as the average grayscale value, maximum and minimum grayscale values ​​of the entire image. The comprehensive grayscale compensation value is intended to provide a reference for the subsequent grayscale compensation processing; by processing the grayscale value of the pixel block of the fingerprint image to be detected and the comprehensive grayscale compensation value, the second grayscale compensation amount corresponding to the grayscale compensation processing of the image to be detected is obtained. This process usually involves some mathematical operations, such as addition, subtraction, multiplication, etc., to realize the correction of the grayscale value of each pixel; the pixel value of the fingerprint boundary pixel block of the fingerprint image to be detected is adjusted to the second grayscale compensation amount.

[0122] The effect of the above technical solution is: by performing grayscale compensation processing on the fingerprint edge pixel block of the fingerprint image to be detected, the grayscale value of the edge area can be extracted, thereby enhancing the information of the fingerprint edge, making the boundary of the target fingerprint image to be detected clearer and more accurate; by extracting the comprehensive grayscale compensation value and combining the grayscale value of the pixel block of the fingerprint image to be detected, the second grayscale compensation amount corresponding to the grayscale compensation processing at the end of the image to be detected can be calculated. In this way, the local grayscale distribution of the fingerprint image can be balanced, so that the brightness difference of the entire image is smaller and has a better visual effect; the pixel value of the fingerprint boundary pixel block of the fingerprint image to be detected is adjusted to the second grayscale compensation amount, which can effectively correct the grayscale difference of the edge area, further enhance the quality and clarity of the target fingerprint image to be detected, and provide more accurate input for subsequent fingerprint recognition. The above formula combines the grayscale value of the pixel block of the fingerprint image to be detected, the comprehensive grayscale compensation value and the adjustment coefficient, which can personalize the grayscale compensation, adjust the grayscale distribution, distinguish the background from the fingerprint, and flexibly control the grayscale compensation effect through the adjustment coefficient, thereby improving the quality and reliability of the target fingerprint image to be detected. At the same time, the above formula is based on the grayscale value of the pixel block of the fingerprint image to be detected. Hti in the formula represents the grayscale value of each pixel block. By summing all pixel blocks and dividing by the number of pixel blocks t, the average grayscale value can be obtained. This can more accurately reflect the grayscale characteristics of the entire fingerprint image, thereby realizing personalized grayscale compensation; the comprehensive grayscale compensation value H z In the formula, it plays the role of adjusting the grayscale distribution. By comprehensively considering the local grayscale characteristics and the overall grayscale compensation value of the fingerprint image to be detected, the grayscale of different areas can be appropriately adjusted according to the specific situation, thereby improving the visual effect of the entire image; C3 and C4 in the formula represent the number of pixel blocks in the fingerprint part and the background part of the fingerprint image to be detected, respectively. By considering the influence of the background part in the calculation, the fingerprint and the background can be better distinguished, and the clarity and recognizability of the target fingerprint image to be detected can be further improved; the adjustment coefficient λ2 plays the role of adjusting the grayscale compensation amount in the formula. By adjusting the value of the coefficient, the intensity and effect of the grayscale compensation can be flexibly controlled to meet different practical application requirements.

[0123] Embodiment 6,

[0124] This embodiment provides a fingerprint image recognition system based on deep learning, and the fingerprint image recognition system includes:

[0125] Data processing module: Perform grayscale compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data in the database, and use the fingerprint sample data after grayscale compensation processing and the original fingerprint sample data deep learning model to perform model training twice, and complete the trained deep learning model; specifically, perform model training twice on the deep learning model by combining the two image pixel conditions of pixel block enlargement and pixel block reduction of the fingerprint sample data in the database with the fingerprint boundary pixel blocks of the fingerprint sample data to perform grayscale compensation processing, and complete the trained deep learning model;

[0126] Image acquisition module: performs grayscale compensation processing on the fingerprint boundary pixel block of the registered user's fingerprint reference image, obtains the fingerprint reference image after grayscale compensation processing, and obtains the fingerprint features of the fingerprint reference image through the deep learning model as the target fingerprint features;

[0127] Image forming module: when the user's fingerprint image to be detected is acquired, the fingerprint edge pixel block of the fingerprint image to be detected is subjected to grayscale compensation processing at the end of the image to be detected to form a target fingerprint image to be detected;

[0128] Result acquisition module: input the target fingerprint image to be detected into the deep learning model, the deep learning model obtains the target fingerprint image to be detected for fingerprint feature recognition, obtains the fingerprint features of the target fingerprint image to be detected, and compares the fingerprint features of the target fingerprint image to be detected with the target fingerprint features to obtain the fingerprint image recognition result.

[0129] The working principle of the above technical solution is: grayscale compensation is performed on the fingerprint sample data in the database, and the processed data and the original data are used to train the deep learning model twice. Here, two different image pixel conditions (enlargement and reduction) are used in combination with the fingerprint boundary pixel blocks for grayscale compensation to increase the robustness and generalization ability of the model; the fingerprint reference image that has been recorded by the user is subjected to grayscale compensation of the fingerprint boundary pixel blocks to obtain the fingerprint reference image after grayscale compensation, and the fingerprint features are extracted as the target fingerprint features through the trained deep learning model; when the user's fingerprint image to be detected is obtained, the fingerprint edge pixel blocks are subjected to grayscale compensation to form the target fingerprint image to be detected; the target fingerprint image to be detected is input into the deep learning model, and the model extracts the fingerprint features of the target fingerprint image to be detected, and compares it with the target fingerprint features to obtain the recognition result of the fingerprint image.

[0130] The effects of the above technical solution are as follows: by grayscale compensation processing of fingerprint sample data and using two deep learning model trainings, the model can better adapt to changes in different lighting, angles and quality, improve the robustness of fingerprint recognition, and effectively cope with complex conditions in real scenes; use the deep learning model to extract fingerprint features, and obtain fingerprint image recognition results by comparing the fingerprint features of the target fingerprint image to be detected with the target fingerprint features. The deep learning model can learn more abstract and advanced feature representations, so it can improve the accuracy of fingerprint image recognition; through two different image pixel conditions (enlargement and reduction) combined with the fingerprint boundary pixel block for grayscale compensation processing, the generalization ability of the model is increased, so that the model can better adapt to fingerprint images of different sizes and resolutions, and improve the generalization ability of fingerprint recognition; based on the deep learning model, it has strong practicality and can be applied to various fingerprint recognition scenarios, such as access control systems, mobile phone unlocking, etc., to provide users with a convenient and safe fingerprint recognition experience. At the same time, traditional fingerprint recognition methods usually require tedious feature engineering, while deep learning models can automatically learn the features of fingerprint images, thereby reducing the difficulty and complexity of feature engineering; deep learning models can automatically adjust model parameters to optimize the performance and accuracy of the model, avoiding the trouble and uncertainty of manually adjusting parameters; and can quickly perform fingerprint recognition, and can process a large amount of fingerprint image data in a short time, making fingerprint recognition more real-time and efficient; the training and reasoning of deep learning models can be performed on general computer hardware, without the need for special fingerprint recognition equipment, thereby reducing the cost and threshold of fingerprint recognition; deep learning models can be trained and optimized for different user groups to adapt to the needs and characteristics of different user groups, thereby achieving personalized customization. By reducing the false recognition rate and improving the anti-interference ability, the overall security of the fingerprint recognition system is improved and the possibility of illegal access is reduced.

[0131] Embodiment 7,

[0132] In this embodiment, the data processing module includes:

[0133] Data extraction module: extracting fingerprint sample data from the database, and using the fingerprint sample data to initially train a deep learning model for fingerprint image feature recognition, to obtain an initial deep learning model after the initial training; wherein the deep learning model adopts a model with a convolutional neural network structure;

[0134] Compensation value acquisition module: for the fingerprint sample data used for initial training, pixel block enlargement and pixel block reduction are performed, and the comprehensive gray compensation value corresponding to the fingerprint sample data used for initial training is obtained by enlarging and reducing the pixel block of the fingerprint sample data;

[0135] Data acquisition module: grayscale compensation processing is performed on the fingerprint boundary pixel block of the fingerprint sample data for initial training by using the comprehensive grayscale compensation value to obtain adjusted sample data; specifically, the comprehensive grayscale compensation value is obtained by using the first grayscale compensation value and the second grayscale compensation value, and the comprehensive data compensation value is used to perform grayscale compensation processing on the fingerprint boundary pixel block of the fingerprint sample data for initial training, so that the grayscale value of the fingerprint boundary pixel block reaches the comprehensive grayscale compensation value;

[0136] Model acquisition module: using the adjusted sample data to perform secondary training on the deep learning model to obtain a trained deep learning model;

[0137] Accuracy acquisition module: using fingerprint verification samples to verify the deep learning model, and obtaining the accuracy of fingerprint image feature recognition of the deep learning model;

[0138] Threshold judgment module: When the accuracy rate is lower than the preset accuracy rate threshold, the deep learning model is optimized until the verified accuracy rate reaches or exceeds the preset accuracy rate threshold.

[0139] The working principle of the above technical solution is: extracting fingerprint sample data from a database, and using the fingerprint sample data to perform initial training on a deep learning model used for fingerprint image feature recognition, to obtain an initial deep learning model after completing the initial training; wherein the deep learning model adopts a model with a convolutional neural network structure; performing pixel block enlargement and pixel block reduction on the fingerprint sample data used for initial training, and obtaining a comprehensive grayscale compensation value corresponding to the fingerprint sample data used for initial training by performing pixel block enlargement and reduction on the fingerprint sample data; performing grayscale compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data used for initial training by using the comprehensive grayscale compensation value to obtain adjusted sample data; specifically The first grayscale compensation value and the second grayscale compensation value are used to obtain a comprehensive grayscale compensation value, and the comprehensive data compensation value is used to perform grayscale compensation processing on the fingerprint boundary pixel block of the fingerprint sample data used for initial training, so that the grayscale value of the fingerprint boundary pixel block reaches the comprehensive grayscale compensation value; the deep learning model is trained again using the adjusted sample data to obtain a deep learning model after training; the deep learning model is verified using a fingerprint verification sample to obtain the accuracy of fingerprint image feature recognition of the deep learning model; when the accuracy is lower than a preset accuracy threshold, the deep learning model is optimized until the accuracy after verification reaches or exceeds the preset accuracy threshold.

[0140] The effects of the above technical solution are as follows: by performing grayscale compensation processing on the fingerprint boundary pixel blocks, the grayscale value of the fingerprint boundary pixel blocks can reach the comprehensive grayscale compensation value, so as to better adapt to the characteristics of the fingerprint image during the deep learning model training process, and improve the accuracy of fingerprint image feature recognition; by performing comprehensive grayscale compensation processing on the fingerprint sample data, the robustness of the model to factors such as different lighting conditions and fingerprint quality differences can be enhanced, and the stability and reliability of the fingerprint recognition system in practical applications can be improved; the technical solution includes a monitoring and optimization mechanism for the model verification accuracy. When the verified accuracy is lower than the preset threshold, the deep learning model can be optimized to adapt it to the characteristics of different data sets, thereby improving the applicability and performance of the system in practical applications; through secondary model training, the model can better adapt to the adjusted sample data, improve the generalization ability of the model, and enable it to achieve better recognition effects when processing new fingerprint sample data. By enlarging and reducing the pixel blocks of the fingerprint sample data and using the comprehensive grayscale compensation value to perform grayscale compensation processing on the fingerprint boundary pixel blocks, the clarity and contrast of the fingerprint image can be enhanced, thereby improving the accuracy of the deep learning model in extracting and identifying fingerprint features; by performing grayscale compensation processing on the fingerprint image to be detected, the impact of problems such as poor fingerprint image quality and insufficient clarity on the fingerprint recognition results can be reduced, thereby improving the security and reliability of fingerprint recognition; using a deep learning model to extract and identify fingerprint features can greatly improve the recognition efficiency and speed, making fingerprint recognition more convenient and efficient. When the accuracy of the deep learning model is lower than the preset accuracy threshold, it can be automatically optimized until the accuracy after verification reaches or exceeds the preset accuracy threshold, thereby ensuring the performance and accuracy of the deep learning model.

[0141] Embodiment 8,

[0142] In this embodiment, the compensation value obtaining module includes:

[0143] The first sample data module: expands the pixel blocks of the fingerprint sample data used for initial training, and obtains the fingerprint sample data after the pixel blocks are expanded as the first sample data; wherein the size range of the expanded pixel blocks is 1.5D-3D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training;

[0144] The first grayscale compensation value module is used to obtain a first grayscale compensation value according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the first sample data; wherein the first grayscale compensation value is obtained by the following formula:

[0145] H xa =ln(1+H x2 -H x1 )

[0146]

[0147]

[0148] Among them, H xa Represents the first grayscale compensation value; H x1 and H x2 denote the first compensation component and the second compensation component respectively; m denotes the number of pixel blocks after the size of the pixel block is enlarged; n denotes the number of pixel blocks before the size of the pixel block is enlarged; H di H represents the grayscale value of the i-th pixel block after the pixel block size is enlarged; yi H represents the grayscale value of the i-th pixel block before the size of the pixel block is enlarged; b Represents the average gray value of the pixel block in the background part; sqrt represents square root operation;

[0149] The second sample data module: performs pixel block reduction on the fingerprint sample data used for initial training, and obtains the fingerprint sample data after the pixel block reduction as the second sample data; wherein the size range of the pixel block reduction is 0.1D-0.5D, and D represents the size of the pixel block of the fingerprint sample data used for initial training;

[0150] The second grayscale compensation value module is used to obtain a second grayscale compensation value according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the second sample data; wherein the second grayscale compensation value is obtained by the following formula:

[0151] H xb =ln(1+H x3 -H x2 )

[0152]

[0153] Among them, H xb Represents the second grayscale compensation value; H x3 represents the third compensation component; k represents the number of pixel blocks after the size of the pixel block is reduced; H si Represents the grayscale value of the i-th pixel block after the size of the pixel block is reduced;

[0154] Comprehensive grayscale compensation value module: uses the first grayscale compensation value and the second grayscale compensation value to obtain a comprehensive grayscale compensation value.

[0155] The comprehensive grayscale compensation value is obtained by the following formula:

[0156]

[0157] Among them, H zIndicates the comprehensive grayscale compensation value; H p Represents the average grayscale value of the fingerprint part of each fingerprint sample data.

[0158] The working principle of the above technical solution is as follows: for the fingerprint sample data used for initial training, pixel blocks are enlarged to obtain the fingerprint sample data after the pixel blocks are enlarged as the first sample data; wherein the size range of the enlarged pixel blocks is 1.5D-3D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training; a first grayscale compensation value is obtained according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the first sample data; pixel blocks are reduced to obtain the fingerprint sample data after the pixel blocks are reduced as the second sample data; wherein the size range of the reduced pixel blocks is 0.1D-0.5D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training; a second grayscale compensation value is obtained according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the second sample data; The fingerprint sample data for initial training is reduced in pixel blocks to obtain the fingerprint sample data after the pixel blocks are reduced as the second sample data; wherein the size range of the reduced pixel blocks is 0.1D-0.5D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training; the second grayscale compensation value is obtained according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the second sample data; the fingerprint sample data for initial training is reduced in pixel blocks to obtain the fingerprint sample data after the pixel blocks are reduced as the second sample data; wherein the size range of the reduced pixel blocks is 0.1D-0.5D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training; the second grayscale compensation value is obtained according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the second sample data; the comprehensive grayscale compensation value is obtained by using the first grayscale compensation value and the second grayscale compensation value. The working principle of the above technical solution is:

[0159] The effect of the above technical solution is: the comprehensive grayscale compensation value is obtained by changing the pixel block size, which can effectively improve the accuracy of the comprehensive grayscale compensation value. By expanding and reducing the pixel block of the fingerprint sample data, the fingerprint image information at different scales can be obtained, so that the system can more comprehensively capture the characteristics and details of the fingerprint image; by calculating the first grayscale compensation value and the second grayscale compensation value, the grayscale difference between the fingerprint part and the background part at different scales is compensated, so as to more accurately describe the grayscale characteristics of the fingerprint image at different scales; using the first grayscale compensation value and the second grayscale compensation value, the comprehensive grayscale compensation value is calculated, which can comprehensively consider the grayscale characteristics of the fingerprint image at different scales, and improve the adaptability of the fingerprint recognition system to fingerprint images of various scales; by comprehensively utilizing multi-scale information and grayscale compensation values, the fingerprint recognition system has stronger robustness and accuracy when facing factors such as different scales and lighting conditions, and improves the practical application value of the system; by performing multi-scale processing on the initial training fingerprint sample data, more comprehensive and diversified data support can be provided for model training, which helps to improve the generalization ability and accuracy of the fingerprint recognition system. The above-mentioned first grayscale compensation value formula takes into account multi-scale factors and grayscale value differences, comprehensively considers the grayscale characteristics of the fingerprint part and the background part, enhances the processing capability and accuracy of multi-scale fingerprint images, and helps to improve the robustness and accuracy of the fingerprint recognition system. At the same time, the first compensation component and the second compensation component in the formula respectively consider the grayscale value difference before and after the pixel block size is enlarged. By comparing the grayscale values ​​of pixel blocks of different sizes, the grayscale changes of the fingerprint part and the background part are more comprehensively described; the m and n parameters in the formula respectively represent the number of pixel blocks after the pixel block size is enlarged and reduced. By considering the grayscale values ​​of pixel blocks of different scales, the grayscale compensation value can fully integrate the fingerprint image information at different scales, thereby improving the system's processing capability for multi-scale fingerprint images; the H in the formula b The parameter represents the average grayscale value of the pixel block in the background part. By introducing the grayscale value of the background part into the calculation, the grayscale features of the fingerprint part and the background part can be better distinguished, and the accuracy of the grayscale compensation value is enhanced. The sqrt operation in the formula is used to amplify and balance the grayscale value difference, so that the grayscale compensation value can better reflect the grayscale changes at different scales, and improve the sensitivity to the grayscale features of the fingerprint image. The above second grayscale compensation value calculation formula comprehensively considers multi-scale factors and grayscale value differences. By processing the grayscale values ​​of pixel blocks at different scales, it enhances the processing capability and accuracy of multi-scale fingerprint images, which helps to improve the robustness and accuracy of the fingerprint recognition system. At the same time, the third compensation component H in the formula x3The grayscale value difference after the pixel block size is reduced is comprehensively considered. By processing the grayscale value of the reduced-size pixel block, the grayscale change of the fingerprint part and the background part is more comprehensively described. The k parameter in the formula represents the number of pixel blocks after the pixel block size is reduced. By considering the grayscale values ​​of pixel blocks of different scales, the second grayscale compensation value can fully integrate the fingerprint image information at different scales, thereby improving the system's processing ability for multi-scale fingerprint images. The H in the formula si The parameter represents the grayscale value of the ith pixel block after the pixel block size is reduced. By processing the grayscale value of the reduced-size pixel block, the second grayscale compensation value can better reflect the grayscale changes at different scales, thereby enhancing the sensitivity to the grayscale features of the fingerprint image. The above-mentioned comprehensive grayscale compensation value formula comprehensively considers the grayscale features of the fingerprint sample data, globally processes the grayscale information of the fingerprint image, and reduces the impact of individual samples on the system, which helps to improve the fingerprint recognition system's processing capability and accuracy for different fingerprint samples and enhances the robustness of the system. At the same time, H in the formula p The parameter represents the average grayscale value of the fingerprint part of each fingerprint sample data. By introducing the average grayscale value of each fingerprint sample data, the comprehensive grayscale compensation value can take into account the grayscale feature differences between different fingerprint samples, thereby enhancing the processing capability of different fingerprint samples. z By processing the average grayscale value of the fingerprint part of each fingerprint sample data, the grayscale characteristics of the entire fingerprint image are comprehensively considered, so that the comprehensive grayscale compensation value can more comprehensively describe the grayscale changes between the fingerprint part and the background part. By introducing the average grayscale value of each fingerprint sample data, the formula can better process the grayscale characteristics of individual samples, reduce the impact of individual samples on the entire system, and improve the stability and robustness of the system.

[0160] Embodiment 9,

[0161] In this embodiment, the image acquisition module includes:

[0162] Gray value extraction module: extracts the gray value of the pixel block of the recorded user's fingerprint reference image;

[0163] Compensation value extraction module: extract comprehensive grayscale compensation value;

[0164] Grayscale compensation value acquisition module: obtains the first grayscale compensation amount corresponding to the grayscale compensation processing of the fingerprint reference image end through the grayscale value of the pixel block of the fingerprint reference image and the comprehensive grayscale compensation value; wherein the first grayscale compensation amount is obtained by the following formula:

[0165]

[0166]

[0167] Among them, H 01 represents the first grayscale compensation amount; r represents the number of pixel blocks of the fingerprint reference image; Hri represents the grayscale value of the i-th pixel block of the fingerprint reference image; H z represents the comprehensive grayscale compensation value; C1 represents the number of pixel blocks of the fingerprint part of the fingerprint reference image; C2 represents the number of pixel blocks of the background part of the fingerprint reference image; λ1 represents the first adjustment coefficient;

[0168] The pixel block adjustment module is used to adjust the pixel values ​​of the fingerprint boundary pixel blocks of the fingerprint reference image to the first grayscale compensation amount to form a fingerprint reference image after grayscale compensation processing.

[0169] The working principle of the above technical solution is: extract the grayscale value of the pixel block of the fingerprint reference image that has been recorded by the user; the system will process the fingerprint reference image that has been recorded by the user, divide the image into different pixel blocks, and extract the grayscale value of each pixel block. These grayscale values ​​reflect the brightness or grayscale level of each area in the fingerprint image; the extraction of the comprehensive grayscale compensation value may involve analyzing the entire fingerprint reference image to determine the overall grayscale characteristics and compensation requirements. This value may be calculated based on factors such as the overall brightness and contrast of the image, and is used for subsequent grayscale compensation processing. By combining the grayscale value of the pixel block of the fingerprint reference image and the comprehensive grayscale compensation value, the system calculates a first grayscale compensation amount, which is usually used for grayscale adjustment of the entire image to achieve overall grayscale compensation processing. Finally, the system adjusts the pixel value of the fingerprint boundary pixel block of the fingerprint reference image to the first grayscale compensation amount to form a fingerprint reference image after grayscale compensation processing. This step is equivalent to fine grayscale adjustment of the local area of ​​the image to enhance the characteristics of the fingerprint and improve the quality of the image.

[0170] The effects of the above technical solution are: by performing grayscale compensation processing on the fingerprint reference image, the interference caused by external factors such as lighting conditions and influences can be effectively reduced, thereby improving the accuracy and robustness of fingerprint recognition; grayscale compensation processing can help highlight and enhance the features in the fingerprint image, making the fingerprint image more distinguishable, thereby improving the reliability of fingerprint recognition; by adjusting the pixel value of the fingerprint reference image, the grayscale compensation processing helps to optimize the quality of the fingerprint image, making the image clearer, easier to analyze and identify; the extraction of the comprehensive grayscale compensation value and the adjustment of the grayscale compensation amount can be adjusted according to different environmental conditions and needs, so that the fingerprint recognition system can better adapt to different working scenarios. The above-mentioned first grayscale compensation amount calculation formula can make the calculation of the grayscale compensation amount more intelligent and personalized, and can better adapt to the characteristics of different fingerprint images, thereby improving the performance of the entire fingerprint recognition system. At the same time, the above formula can be personalized according to the characteristics of the specific fingerprint reference image. By considering the number of pixel blocks in the fingerprint part and the background part, the formula can compensate the grayscale features of different parts to different degrees to adapt to the characteristics of different fingerprint images. The formula contains the comprehensive grayscale compensation value H z , which enables the calculation of grayscale compensation to comprehensively consider the grayscale features of the entire fingerprint image, rather than just being limited to the grayscale values ​​of the local area. This allows for a more comprehensive adjustment of the image's grayscale, improving the quality of the fingerprint image. The introduction of the adjustment coefficient λ1 makes the grayscale compensation calculation in the formula adjustable, and the parameters can be adjusted according to actual conditions to obtain the best grayscale compensation effect. Through this personalized, comprehensively considered and adjustable grayscale compensation calculation, the grayscale of the fingerprint reference image can be adjusted more accurately, thereby improving the accuracy and robustness of the fingerprint recognition system for fingerprint images.

[0171] Embodiment 10

[0172] In this embodiment, the image forming module includes:

[0173] Detection module: extracts the gray value of the pixel block of the fingerprint image to be detected;

[0174] Extraction module: extract comprehensive grayscale compensation value;

[0175] Acquisition module: Acquisition of a second grayscale compensation amount corresponding to the grayscale compensation processing of the image to be detected through the grayscale value of the pixel block of the fingerprint image to be detected and the comprehensive grayscale compensation value; wherein the second grayscale compensation amount is obtained by the following formula:

[0176]

[0177]

[0178] Among them, H02 represents the second grayscale compensation amount; t represents the number of pixel blocks of the fingerprint image to be detected; Hti represents the grayscale value of the i-th pixel block of the fingerprint image to be detected; H z represents the comprehensive grayscale compensation value; C3 represents the number of pixel blocks of the fingerprint part of the fingerprint image to be detected; C4 represents the number of pixel blocks of the background part of the fingerprint image to be detected; λ2 represents the second adjustment coefficient.

[0179] Image forming module: adjusts the pixel value of the fingerprint boundary pixel block of the fingerprint image to be detected to the second grayscale compensation amount to form the target fingerprint image to be detected.

[0180] The working principle of the above technical solution is: the fingerprint image to be detected is segmented into pixel blocks, each pixel block contains a certain number of pixels. Then, the grayscale values ​​of these pixel blocks are extracted. The grayscale value represents the color information of each pixel, ranging from 0 to 255, where 0 represents black and 255 represents white. This process is to obtain the brightness information of each pixel in the image; the extraction of the comprehensive grayscale compensation value is to consider the brightness and contrast of the entire image. It is obtained by processing various statistical features such as the average grayscale value, maximum and minimum grayscale values ​​of the entire image. The comprehensive grayscale compensation value is intended to provide a reference for the subsequent grayscale compensation processing; by processing the grayscale value of the pixel block of the fingerprint image to be detected and the comprehensive grayscale compensation value, the second grayscale compensation amount corresponding to the grayscale compensation processing of the image to be detected is obtained. This process usually involves some mathematical operations, such as addition, subtraction, multiplication, etc., to realize the correction of the grayscale value of each pixel; the pixel value of the fingerprint boundary pixel block of the fingerprint image to be detected is adjusted to the second grayscale compensation amount.

[0181] The effect of the above technical solution is: by performing grayscale compensation processing on the fingerprint edge pixel block of the fingerprint image to be detected, the grayscale value of the edge area can be extracted, thereby enhancing the information of the fingerprint edge, making the boundary of the target fingerprint image to be detected clearer and more accurate; by extracting the comprehensive grayscale compensation value and combining the grayscale value of the pixel block of the fingerprint image to be detected, the second grayscale compensation amount corresponding to the grayscale compensation processing at the end of the image to be detected can be calculated. In this way, the local grayscale distribution of the fingerprint image can be balanced, so that the brightness difference of the entire image is smaller and has a better visual effect; the pixel value of the fingerprint boundary pixel block of the fingerprint image to be detected is adjusted to the second grayscale compensation amount, which can effectively correct the grayscale difference of the edge area, further enhance the quality and clarity of the target fingerprint image to be detected, and provide more accurate input for subsequent fingerprint recognition. The above formula combines the grayscale value of the pixel block of the fingerprint image to be detected, the comprehensive grayscale compensation value and the adjustment coefficient, which can personalize the grayscale compensation, adjust the grayscale distribution, distinguish the background from the fingerprint, and flexibly control the grayscale compensation effect through the adjustment coefficient, thereby improving the quality and reliability of the target fingerprint image to be detected. At the same time, the above formula is based on the grayscale value of the pixel block of the fingerprint image to be detected. Hti in the formula represents the grayscale value of each pixel block. By summing all pixel blocks and dividing by the number of pixel blocks t, the average grayscale value can be obtained. This can more accurately reflect the grayscale characteristics of the entire fingerprint image, thereby realizing personalized grayscale compensation; the comprehensive grayscale compensation value H z In the formula, it plays the role of adjusting the grayscale distribution. By comprehensively considering the local grayscale characteristics and the overall grayscale compensation value of the fingerprint image to be detected, the grayscale of different areas can be appropriately adjusted according to the specific situation, thereby improving the visual effect of the entire image; C3 and C4 in the formula represent the number of pixel blocks in the fingerprint part and the background part of the fingerprint image to be detected, respectively. By considering the influence of the background part in the calculation, the fingerprint and the background can be better distinguished, and the clarity and recognizability of the target fingerprint image to be detected can be further improved; the adjustment coefficient λ2 plays the role of adjusting the grayscale compensation amount in the formula. By adjusting the value of the coefficient, the intensity and effect of the grayscale compensation can be flexibly controlled to meet different practical application requirements.

[0182] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A fingerprint image recognition method based on deep learning, characterized in that: The fingerprint image recognition method comprises: Perform grayscale compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data in the database, and use the fingerprint sample data after grayscale compensation processing and the original fingerprint sample data deep learning model to perform model training twice to complete the trained deep learning model; Performing grayscale compensation processing on the fingerprint boundary pixel block of the registered user's fingerprint reference image to obtain the fingerprint reference image after the grayscale compensation processing, and obtaining the fingerprint feature of the fingerprint reference image through the deep learning model as the target fingerprint feature; When the user's fingerprint image to be detected is obtained, the fingerprint edge pixel blocks of the fingerprint image to be detected are subjected to grayscale compensation processing at the image end to be detected to form a target fingerprint image to be detected; The target fingerprint image to be detected is input into the deep learning model, the deep learning model obtains the target fingerprint image to be detected for fingerprint feature recognition, obtains the fingerprint features of the target fingerprint image to be detected, and compares the image fingerprint features of the target fingerprint to be detected with the target fingerprint features to obtain the fingerprint image recognition result; The method of performing grayscale compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data in the database, and performing two model trainings using the fingerprint sample data after grayscale compensation processing and the original fingerprint sample data deep learning model to complete the trained deep learning model includes: Extracting fingerprint sample data from a database, and using the fingerprint sample data to initially train a deep learning model for fingerprint image feature recognition, to obtain an initial deep learning model after the initial training is completed; Performing pixel block enlargement and pixel block reduction on the fingerprint sample data used for initial training, and obtaining a comprehensive grayscale compensation value corresponding to the fingerprint sample data used for initial training by performing pixel block enlargement and pixel block reduction on the fingerprint sample data; Performing grayscale compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data used for initial training by using the comprehensive grayscale compensation value to obtain adjusted sample data; Performing secondary training on the deep learning model using the adjusted sample data to obtain a trained deep learning model; Using fingerprint verification samples to verify the deep learning model, and obtaining the accuracy of fingerprint image feature recognition of the deep learning model; When the accuracy is lower than a preset accuracy threshold, the deep learning model is optimized until the verified accuracy reaches or exceeds the preset accuracy threshold.

2. The fingerprint image recognition method based on deep learning according to claim 1, characterized in that: The pixel blocks are enlarged and reduced for the fingerprint sample data used for the initial training, and the comprehensive grayscale compensation value corresponding to the fingerprint sample data used for the initial training is obtained by enlarging and reducing the pixel blocks of the fingerprint sample data, including: Expanding the pixel blocks of the fingerprint sample data used for initial training to obtain the fingerprint sample data after the pixel blocks are expanded as the first sample data; wherein the size range of the expanded pixel blocks is 1.5D-3D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training; Obtaining a first grayscale compensation value according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the first sample data; The fingerprint sample data used for initial training is reduced in pixel blocks to obtain the fingerprint sample data after the pixel blocks are reduced as the second sample data; wherein the size range of the reduced pixel blocks is 0.1D-0.5D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training; Obtaining a second grayscale compensation value according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the second sample data; A comprehensive grayscale compensation value is obtained by using the first grayscale compensation value and the second grayscale compensation value.

3. The fingerprint image recognition method based on deep learning according to claim 1, characterized in that: Performing grayscale compensation processing on the fingerprint reference image side of the fingerprint boundary pixel block of the registered user's fingerprint reference image to obtain the fingerprint reference image after grayscale compensation processing, including: Extracting the grayscale value of the pixel block of the recorded user's fingerprint reference image; Extracting comprehensive grayscale compensation value; Obtaining a first grayscale compensation amount corresponding to the grayscale compensation processing of the fingerprint reference image end through the grayscale value of the pixel block of the fingerprint reference image and the comprehensive grayscale compensation value; The pixel values ​​of the fingerprint boundary pixel blocks of the fingerprint reference image are adjusted to a first grayscale compensation amount to form a fingerprint reference image after grayscale compensation processing.

4. The fingerprint image recognition method based on deep learning according to claim 1, characterized in that: When the user's fingerprint image to be detected is obtained, the grayscale compensation processing of the fingerprint edge pixel block of the fingerprint image to be detected is performed on the image end to be detected to form a target fingerprint image to be detected, including: Extracting the grayscale value of the pixel block of the fingerprint image to be detected; Extracting comprehensive grayscale compensation value; Obtaining a second grayscale compensation amount corresponding to the grayscale compensation processing of the image to be detected through the grayscale value of the pixel block of the fingerprint image to be detected and the comprehensive grayscale compensation value; The pixel values ​​of the fingerprint boundary pixel blocks of the fingerprint image to be detected are adjusted to the second grayscale compensation amount to form a target fingerprint image to be detected.

5. A fingerprint image recognition system based on deep learning, characterized in that: The fingerprint image recognition system comprises: Data processing module: Perform grayscale compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data in the database, and use the fingerprint sample data after grayscale compensation processing and the original fingerprint sample data deep learning model to perform two model trainings to complete the trained deep learning model; Image acquisition module: performs grayscale compensation processing on the fingerprint boundary pixel block of the registered user's fingerprint reference image, obtains the fingerprint reference image after grayscale compensation processing, and obtains the fingerprint features of the fingerprint reference image through the deep learning model as the target fingerprint features; Image forming module: when the user's fingerprint image to be detected is acquired, the fingerprint edge pixel block of the fingerprint image to be detected is subjected to grayscale compensation processing at the end of the image to be detected to form a target fingerprint image to be detected; Result acquisition module: input the target fingerprint image to be detected into the deep learning model, the deep learning model acquires the target fingerprint image to be detected for fingerprint feature recognition, obtains the fingerprint features of the target fingerprint image to be detected, and compares the fingerprint features of the target fingerprint image to be detected with the target fingerprint features to obtain the fingerprint image recognition result; The data processing module comprises: Data extraction module: extracts fingerprint sample data from the database, and uses the fingerprint sample data to perform initial training on a deep learning model used for fingerprint image feature recognition, thereby obtaining an initial deep learning model after the initial training is completed; Compensation value acquisition module: for the fingerprint sample data used for initial training, pixel block enlargement and pixel block reduction are performed, and the comprehensive gray compensation value corresponding to the fingerprint sample data used for initial training is obtained by enlarging and reducing the pixel block of the fingerprint sample data; Data acquisition module: performing grayscale compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data used for initial training by using the comprehensive grayscale compensation value to obtain adjusted sample data; Model acquisition module: using the adjusted sample data to perform secondary training on the deep learning model to obtain a trained deep learning model; Accuracy acquisition module: using fingerprint verification samples to verify the deep learning model, and obtaining the accuracy of fingerprint image feature recognition of the deep learning model; Threshold judgment module: When the accuracy rate is lower than the preset accuracy rate threshold, the deep learning model is optimized until the verified accuracy rate reaches or exceeds the preset accuracy rate threshold.

6. The fingerprint image recognition system based on deep learning according to claim 5, characterized in that: The compensation value obtaining module comprises: The first sample data module: expands the pixel blocks of the fingerprint sample data used for initial training, and obtains the fingerprint sample data after the pixel blocks are expanded as the first sample data; wherein the size range of the expanded pixel blocks is 1.5D-3D, and D represents the size of the pixel blocks of the fingerprint sample data used for initial training; A first grayscale compensation value module: obtaining a first grayscale compensation value according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the first sample data; The second sample data module: performs pixel block reduction on the fingerprint sample data used for initial training, and obtains the fingerprint sample data after the pixel block reduction as the second sample data; wherein the size range of the pixel block reduction is 0.1D-0.5D, and D represents the size of the pixel block of the fingerprint sample data used for initial training; A second grayscale compensation value module: obtaining a second grayscale compensation value according to the grayscale values ​​of the pixel blocks of the fingerprint part and the background part in the second sample data; Comprehensive grayscale compensation value module: uses the first grayscale compensation value and the second grayscale compensation value to obtain a comprehensive grayscale compensation value.

7. The fingerprint image recognition system based on deep learning according to claim 5, characterized in that: The image acquisition module comprises: Gray value extraction module: extracts the gray value of the pixel block of the recorded user's fingerprint reference image; Compensation value extraction module: extract comprehensive grayscale compensation value; Grayscale compensation value acquisition module: acquires a first grayscale compensation amount corresponding to the grayscale compensation processing of the fingerprint reference image end through the grayscale value of the pixel block of the fingerprint reference image and the comprehensive grayscale compensation value; The pixel block adjustment module is used to adjust the pixel values ​​of the fingerprint boundary pixel blocks of the fingerprint reference image to the first grayscale compensation amount to form a fingerprint reference image after grayscale compensation processing.

8. The fingerprint image recognition system based on deep learning according to claim 5, characterized in that: The image forming module comprises: Detection module: extracts the gray value of the pixel block of the fingerprint image to be detected; Extraction module: extract comprehensive grayscale compensation value; An acquisition module: acquiring a second grayscale compensation amount corresponding to the grayscale compensation processing of the image to be detected through the grayscale value of the pixel block of the fingerprint image to be detected and the comprehensive grayscale compensation value; Image forming module: adjusts the pixel value of the fingerprint boundary pixel block of the fingerprint image to be detected to the second grayscale compensation amount to form the target fingerprint image to be detected.

Images