CN117542088A

CN117542088A - Fingerprint image recognition method and system based on deep learning

Info

Publication number: CN117542088A
Application number: CN202311518707.7A
Authority: CN
Inventors: 罗洪昌; 张飞飞
Original assignee: Hangzhou Synochip Data Security Technology Co ltd
Current assignee: Hangzhou Synochip Data Security Technology Co ltd
Priority date: 2023-11-15
Filing date: 2023-11-15
Publication date: 2024-02-09

Abstract

The invention provides a fingerprint image identification method and a fingerprint image identification system based on deep learning. The fingerprint image recognition method comprises the steps of performing model training on the fingerprint sample data subjected to gray level compensation processing and an original fingerprint sample data deep learning model twice, and completing the trained deep learning model; the fingerprint reference image after gray level compensation processing is obtained by carrying out gray level compensation processing on the fingerprint reference image at the fingerprint reference image end, and fingerprint characteristics are obtained; carrying out gray level compensation processing on a fingerprint edge pixel block of a fingerprint image to be detected at the image end to be detected to form a target fingerprint image to be detected; and acquiring a target fingerprint image to be detected through the deep learning model to perform fingerprint feature recognition, acquiring 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 acquire a fingerprint image recognition result. The system comprises modules corresponding to the method steps.

Description

Fingerprint image recognition method and system based on deep learning

Technical Field

The invention relates to the technical field of fingerprint image recognition, in particular to a fingerprint image recognition method and system based on deep learning.

Background

The fingerprint identification technology is one of biological identification technology, and can realize the identification of individuals by comparing fingerprint characteristics with fingerprint information in a database. Along with the development of technology, fingerprint identification technology is widely applied in the fields of security authentication, access control systems and the like. However, in practical applications, due to the diversity of fingerprint samples and different acquisition devices, the acquired fingerprint image often has the problem of uneven gray levels of boundary pixel blocks, which brings challenges to the extraction and identification of fingerprint features.

In order to solve this problem, some studies have proposed a method of performing gray-scale compensation processing on a fingerprint image. These methods improve the gray scale distribution of boundary pixel blocks, mainly by enlarging or reducing the pixel blocks of the fingerprint image. However, these methods tend to be directed to a single pixel block processing manner, and neglecting the processing effect of the pixel block under different conditions.

Furthermore, for deep learning models, the quality and quantity of training data will directly impact the performance of the model. In the prior art, the training data is mainly derived from the registered fingerprint reference images of the user, but the quality and number of these images are limited, and therefore the performance of the model is limited.

Disclosure of Invention

The invention provides a fingerprint image recognition method and a fingerprint image recognition system based on deep learning, which are used for solving the problems that in the prior art, due to sweat stain, water stain or problems of image acquisition equipment, a linkage image is generated between adjacent fingerprint lines, so that the problem that the boundary definition of a fingerprint line is lower is caused, and the accuracy of fingerprint recognition is influenced:

the invention provides a fingerprint image identification method based on deep learning, which comprises the following steps:

s1: carrying out gray level compensation processing on fingerprint sample data in a database by using a fingerprint boundary pixel block, and carrying out model training twice by using the fingerprint sample data subjected to the gray level compensation processing and an original fingerprint sample data deep learning model to finish a trained deep learning model;

s2: gray level compensation processing is carried out on a fingerprint boundary pixel block of the input fingerprint reference image of the user at a fingerprint reference image end, a fingerprint reference image after gray level compensation processing is obtained, and fingerprint characteristics of the fingerprint reference image are obtained through the deep learning model and are used as target fingerprint characteristics;

s3: when a fingerprint image to be detected of a user is obtained, gray level compensation processing of an image end to be detected is carried out on a fingerprint edge pixel block of the fingerprint image to be detected, so that a target fingerprint image to be detected is formed;

S4: and inputting the target fingerprint image to be detected into a deep learning model, acquiring the target fingerprint image to be detected by the deep learning model to perform fingerprint feature recognition, acquiring 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 acquire a fingerprint image recognition result.

Further, the gray level compensation processing of the fingerprint boundary pixel block is performed on the fingerprint sample data in the database, and the two model training is performed by using the fingerprint sample data after the gray level compensation processing and the original fingerprint sample data deep learning model, so as to complete the trained deep learning model, which comprises the following steps:

s11: extracting fingerprint sample data from a database, and utilizing 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;

s12: performing pixel block expansion and pixel block reduction on fingerprint sample data for initial training, and obtaining a comprehensive gray level compensation value corresponding to the fingerprint sample data for initial training by performing pixel block expansion and pixel block reduction on the fingerprint sample data;

S13: performing gray level compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data for initial training through the comprehensive gray level compensation value to obtain adjusted sample data;

s14: performing secondary training on the deep learning model by using the adjusted sample data to obtain a trained deep learning model;

s15: verifying the deep learning model by using a fingerprint verification sample to obtain the accuracy of fingerprint image feature recognition of the deep learning model;

s16: and when the accuracy is lower than a preset accuracy threshold, optimizing the deep learning model until the verified accuracy reaches or exceeds the preset accuracy threshold.

Further, performing pixel block expansion and pixel block reduction on fingerprint sample data for initial training, and obtaining a comprehensive gray compensation value corresponding to the fingerprint sample data for initial training by performing pixel block expansion and pixel block reduction on the fingerprint sample data, including:

s121: performing pixel block expansion on fingerprint sample data for initial training to obtain fingerprint sample data with expanded pixel blocks as first sample data; wherein the pixel block has an enlarged size range of 1.5D-3D, D representing the size of the pixel block of fingerprint sample data for initial training;

S122: obtaining a first gray compensation value according to gray values of pixel blocks of a fingerprint part and a background part in the first sample data;

s123: performing pixel block reduction on fingerprint sample data for initial training to obtain fingerprint sample data with reduced pixel blocks as second sample data; wherein the size range of the pixel block is 0.1D-0.5D, and D represents the size of the pixel block of the fingerprint sample data for initial training;

s124: obtaining a second gray level compensation value according to gray level values of pixel blocks of the fingerprint part and the background part in the second sample data;

s125: and acquiring a comprehensive gray scale compensation value by using the first gray scale compensation value and the second gray scale compensation value.

Further, gray level compensation processing of a fingerprint reference image end is performed on a fingerprint boundary pixel block of the registered fingerprint reference image of the user, and a fingerprint reference image after gray level compensation processing is obtained, including:

s21: extracting gray values of pixel blocks of the registered fingerprint reference image of the user;

s22: extracting a comprehensive gray level compensation value;

s23: acquiring a first gray level compensation amount corresponding to gray level compensation processing of the fingerprint reference image end through the gray level value and the comprehensive gray level compensation value of the pixel block of the fingerprint reference image;

S24: and adjusting the pixel value of the fingerprint boundary pixel block of the fingerprint reference image to be a first gray level compensation amount to form a fingerprint reference image after gray level compensation processing.

Further, when a fingerprint image to be detected of a user is obtained, gray compensation processing of an image end to be detected is performed on a fingerprint edge pixel block of the fingerprint image to be detected, so as to form a target fingerprint image to be detected, including:

s31: extracting gray values of pixel blocks of a fingerprint image to be detected;

s32: extracting a comprehensive gray level compensation value;

s33: acquiring a second gray level compensation amount corresponding to gray level compensation processing of the image end to be detected through the gray level value and the comprehensive gray level compensation value of the pixel block of the fingerprint image to be detected;

s34: and adjusting the pixel value of the fingerprint boundary pixel block of the fingerprint image to be detected to be a second gray level compensation amount to form the target fingerprint image to be detected.

The invention provides a fingerprint image recognition system based on deep learning, which comprises:

and a data processing module: carrying out gray level compensation processing on fingerprint sample data in a database by using a fingerprint boundary pixel block, and carrying out model training twice by using the fingerprint sample data subjected to the gray level compensation processing and an original fingerprint sample data deep learning model to finish a trained deep learning model;

An image acquisition module: gray level compensation processing is carried out on a fingerprint boundary pixel block of the input fingerprint reference image of the user at a fingerprint reference image end, a fingerprint reference image after gray level compensation processing is obtained, and fingerprint characteristics of the fingerprint reference image are obtained through the deep learning model and are used as target fingerprint characteristics;

an image forming module: when a fingerprint image to be detected of a user is obtained, gray level compensation processing of an image end to be detected is carried out on a fingerprint edge pixel block of the fingerprint image to be detected, so that a target fingerprint image to be detected is formed;

the result obtaining module: and inputting the target fingerprint image to be detected into a deep learning model, acquiring the target fingerprint image to be detected by the deep learning model to perform fingerprint feature recognition, acquiring 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 acquire a fingerprint image recognition result.

Further, the data processing module includes:

and a data extraction module: extracting fingerprint sample data from a database, and utilizing 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;

The compensation value obtaining module: performing pixel block expansion and pixel block reduction on fingerprint sample data for initial training, and obtaining a comprehensive gray level compensation value corresponding to the fingerprint sample data for initial training by performing pixel block expansion and pixel block reduction on the fingerprint sample data;

and a data acquisition module: performing gray level compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data for initial training through the comprehensive gray level compensation value to obtain adjusted sample data;

model acquisition module: performing secondary training on the deep learning model by using the adjusted sample data to obtain a trained deep learning model;

the accuracy obtaining module is used for: verifying the deep learning model by using a fingerprint verification sample to obtain the accuracy of fingerprint image feature recognition of the deep learning model;

threshold judgment module: and when the accuracy is lower than a preset accuracy threshold, optimizing the deep learning model until the verified accuracy reaches or exceeds the preset accuracy threshold.

Further, the compensation value obtaining module includes:

a first sample data module: performing pixel block expansion on fingerprint sample data for initial training to obtain fingerprint sample data with expanded pixel blocks as first sample data; wherein the pixel block has an enlarged size range of 1.5D-3D, D representing the size of the pixel block of fingerprint sample data for initial training;

A first gray compensation value module: obtaining a first gray compensation value according to gray values of pixel blocks of a fingerprint part and a background part in the first sample data;

a second sample data module: performing pixel block reduction on fingerprint sample data for initial training to obtain fingerprint sample data with reduced pixel blocks as second sample data; wherein the size range of the pixel block is 0.1D-0.5D, and D represents the size of the pixel block of the fingerprint sample data for initial training;

the second gray level compensation value module: obtaining a second gray level compensation value according to gray level values of pixel blocks of the fingerprint part and the background part in the second sample data;

and a comprehensive gray level compensation value module: and acquiring a comprehensive gray scale compensation value by using the first gray scale compensation value and the second gray scale compensation value.

Further, the image obtaining module includes:

gray value extraction module: extracting gray values of pixel blocks of the registered fingerprint reference image of the user;

and the compensation value extraction module is used for: extracting a comprehensive gray level compensation value;

the gray compensation value acquisition module: acquiring a first gray level compensation amount corresponding to gray level compensation processing of the fingerprint reference image end through the gray level value and the comprehensive gray level compensation value of the pixel block of the fingerprint reference image;

And a pixel block adjusting module: and adjusting the pixel value of the fingerprint boundary pixel block of the fingerprint reference image to be a first gray level compensation amount to form a fingerprint reference image after gray level compensation processing.

Further, the image forming module includes:

and a detection module: extracting gray values of pixel blocks of a fingerprint image to be detected;

and an extraction module: extracting a comprehensive gray level compensation value;

the acquisition module is used for: acquiring a second gray level compensation amount corresponding to gray level compensation processing of the image end to be detected through the gray level value and the comprehensive gray level compensation value of the pixel block of the fingerprint image to be detected;

an image forming module: and adjusting the pixel value of the fingerprint boundary pixel block of the fingerprint image to be detected to be a second gray level compensation amount to form the target fingerprint image to be detected.

The invention has the beneficial effects that: the definition and contrast of the fingerprint image can be enhanced by carrying out gray compensation processing on the fingerprint boundary pixel blocks, so that the accuracy of extracting and identifying fingerprint features by the deep learning model is improved; the fingerprint sample data after gray level compensation processing and the original fingerprint sample data are combined in the training process of the deep learning model, so that the deep learning model has good adaptability and robustness to fingerprint images under different conditions; the influence of the problems of poor quality, insufficient definition and the like of the fingerprint image on the fingerprint identification result can be reduced by carrying out gray level compensation processing on the fingerprint image to be detected, so that the safety and reliability of fingerprint identification are improved; the deep learning model is adopted to extract and identify fingerprint features, so that the identification efficiency and speed can be greatly improved, and the fingerprint identification is more convenient and efficient; the fingerprint identification scheme can be suitable for various scenes needing identity verification, such as mobile phone unlocking, access control systems and the like, and has wide application prospect and market prospect. Meanwhile, through the application of a deep learning algorithm, the system can effectively identify and remove the alliance shadow caused by sweat stain, water stain or image acquisition equipment and other problems, so that the definition and quality of fingerprint images are improved; the fingerprint image with higher definition is beneficial to improving the accuracy and the robustness of a fingerprint identification system, reducing the false recognition rate and increasing the success rate of fingerprint identification; the deep learning method can learn and understand the fingerprint image characteristics under various different conditions, so that the processing capacity of the system for resisting sweat, water stains and other interference factors is improved, and the fingerprint identification can keep high accuracy in a complex environment; through solving the lower problem of fingerprint line boundary definition, the user will obtain more stable, quick and reliable discernment experience when using fingerprint identification system, promotes whole user experience.

Drawings

FIG. 1 is a step diagram of a fingerprint image recognition method based on deep learning according to the present invention;

fig. 2 is a block diagram of a fingerprint image recognition system based on deep learning according to the present invention.

Detailed Description

The preferred embodiments of the present invention will be described below with reference to the accompanying drawings, it being understood that the preferred embodiments described herein are for illustration and explanation of the present invention only, and are not intended to limit the present invention.

Example 1,

In this embodiment, a fingerprint image recognition method based on deep learning includes:

s1: carrying out gray level compensation processing on fingerprint sample data in a database by using a fingerprint boundary pixel block, and carrying out model training twice by using the fingerprint sample data subjected to the gray level compensation processing and an original fingerprint sample data deep learning model to finish a trained deep learning model; specifically, the deep learning model is trained twice by combining two image pixel conditions of enlarging pixel blocks and shrinking pixel blocks of fingerprint sample data in a database and a mode of gray compensation processing of fingerprint boundary pixel blocks of the fingerprint sample data, so that the trained deep learning model is completed;

The working principle of the technical scheme is as follows: and carrying out gray level compensation processing on fingerprint sample data in the database, and carrying out two-time deep learning model training by using the processed data and the original data. Two different image pixel conditions (expansion and contraction) are adopted to combine with a fingerprint boundary pixel block to carry out gray level compensation processing, so that the robustness and generalization capability of the model are improved; gray level compensation processing is carried out on the fingerprint reference image input by the user to obtain a fingerprint reference image after gray level compensation processing, and fingerprint characteristics are extracted through a trained deep learning model to serve as target fingerprint characteristics; when a fingerprint image to be detected of a user is obtained, gray level compensation processing of a fingerprint edge pixel block is carried out on the fingerprint image to be detected, so that a target fingerprint image to be detected is formed; inputting the target fingerprint image to be detected into a deep learning model, extracting fingerprint characteristics of the target fingerprint image to be detected by the model, and comparing the fingerprint characteristics with the target fingerprint characteristics to obtain a recognition result of the fingerprint image.

The technical scheme has the effects that: the fingerprint sample data is subjected to gray level compensation processing, and the two-time deep learning model training is utilized, so that the model can be better adapted to different illumination, angle, quality and other changes, the robustness of fingerprint identification is improved, and the complex conditions in a real scene are effectively met; and extracting fingerprint features by using the deep learning model, and comparing the fingerprint features of the target fingerprint image to be detected with the target fingerprint features to obtain a fingerprint image identification result. The deep learning model can learn more abstract and advanced characteristic representation, so that the accuracy of fingerprint image recognition can be improved; the gray level compensation processing is carried out by combining two different image pixel conditions (expansion and shrinkage) with the fingerprint boundary pixel blocks, so that the generalization capability of the model is improved, the model can be better adapted to fingerprint images with different sizes and resolutions, and the generalization capability of fingerprint identification is improved; based on the deep learning model, the method has strong practicability, can be applied to various fingerprint identification scenes, such as access control systems, mobile phone unlocking and the like, and provides convenient and safe fingerprint identification experience for users. Meanwhile, the traditional fingerprint identification method generally needs to carry out complicated characteristic engineering, and the deep learning model can automatically learn the characteristics of the fingerprint image, so that the difficulty and complexity of the characteristic engineering are reduced; the deep learning model can automatically adjust model parameters to optimize the performance and accuracy of the model, so that the trouble and uncertainty of manually adjusting the parameters are avoided; the fingerprint identification can be rapidly carried out, and a large amount of fingerprint image data can be processed in a short time, so that the fingerprint identification is more real-time and efficient; training and reasoning of the deep learning model can be performed on general computer hardware, and special fingerprint identification equipment is not needed, so that the cost and threshold of fingerprint identification are reduced; the deep learning model can be trained and optimized for different user groups so as to adapt to the requirements and characteristics of the different user groups, thereby realizing the effect of personalized customization. By reducing the false recognition rate, improving the anti-interference capability and the like, the security of the fingerprint identification system is improved as a whole, and the possibility of illegal access is reduced.

EXAMPLE 2,

In this embodiment, gray level compensation processing is performed on fingerprint sample data in a database, and two model training is performed by using the fingerprint sample data after gray level compensation processing and an original fingerprint sample data deep learning model, so as to complete the trained deep learning model, including:

s11: extracting fingerprint sample data from a database, and utilizing 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; the deep learning model adopts a model with a convolutional neural network structure;

s13: performing gray level compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data for initial training through the comprehensive gray level compensation value to obtain adjusted sample data; specifically, the first gray level compensation value and the second gray level compensation value are utilized to obtain a comprehensive gray level compensation value, and the comprehensive data compensation value is utilized to carry out gray level compensation processing on the fingerprint boundary pixel block of the fingerprint sample data for initial training, so that the gray level value of the fingerprint boundary pixel block reaches the comprehensive gray level compensation value;

The working principle of the technical scheme is as follows: extracting fingerprint sample data from a database, and utilizing 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; the deep learning model adopts a model with a convolutional neural network structure; performing pixel block expansion and pixel block reduction on fingerprint sample data for initial training, and obtaining a comprehensive gray level compensation value corresponding to the fingerprint sample data for initial training by performing pixel block expansion and pixel block reduction on the fingerprint sample data; performing gray level compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data for initial training through the comprehensive gray level compensation value to obtain adjusted sample data; specifically, the first gray level compensation value and the second gray level compensation value are utilized to obtain a comprehensive gray level compensation value, and the comprehensive data compensation value is utilized to carry out gray level compensation processing on the fingerprint boundary pixel block of the fingerprint sample data for initial training, so that the gray level value of the fingerprint boundary pixel block reaches the comprehensive gray level compensation value; performing secondary training on the deep learning model by using the adjusted sample data to obtain a trained deep learning model; verifying the deep learning model by using a fingerprint verification sample to obtain the accuracy of fingerprint image feature recognition of the deep learning model; and when the accuracy is lower than a preset accuracy threshold, optimizing the deep learning model until the verified accuracy reaches or exceeds the preset accuracy threshold.

The technical scheme has the effects that: the gray level compensation processing is carried out on the fingerprint boundary pixel blocks, so that the gray level value of the fingerprint boundary pixel blocks can reach the comprehensive gray level compensation value, the characteristics of the fingerprint images are better adapted in the training process of the deep learning model, and the accuracy rate of fingerprint image feature recognition is improved; the robustness of the model to factors such as different illumination conditions, fingerprint quality differences and the like can be enhanced by carrying out comprehensive gray level compensation processing on fingerprint sample data, and the stability and reliability of the fingerprint identification system in practical application are improved; the technical scheme comprises a monitoring and optimizing mechanism for model verification accuracy, and when the verified accuracy is lower than a preset threshold, the deep learning model can be optimized to adapt to the characteristics of different data sets, so that the applicability and performance of the system in practical application are improved; through secondary model training, the model can better adapt to the adjusted sample data, the generalization capability of the model is improved, and a better recognition effect can be obtained when new fingerprint sample data are processed. The definition and contrast of the fingerprint image can be enhanced by expanding and shrinking the pixel blocks of the fingerprint sample data and carrying out gray compensation processing on the pixel blocks of the fingerprint boundary by utilizing the comprehensive gray compensation value, so that the accuracy of extracting and identifying the fingerprint features by the deep learning model is improved; the influence of the problems of poor quality, insufficient definition and the like of the fingerprint image on the fingerprint identification result can be reduced by carrying out gray level compensation processing on the fingerprint image to be detected, so that the safety and reliability of fingerprint identification are improved; the deep learning model is adopted to extract and identify fingerprint features, so that the identification efficiency and speed can be greatly improved, and the fingerprint identification is more convenient and efficient. When the accuracy of the deep learning model is lower than a preset accuracy threshold, the optimization can be automatically performed until the verified accuracy reaches or exceeds the preset accuracy threshold, so that the performance and accuracy of the deep learning model are ensured.

EXAMPLE 3,

In this embodiment, pixel block expansion and pixel block reduction are performed on fingerprint sample data for initial training, and a comprehensive gray compensation value corresponding to the fingerprint sample data for initial training is obtained by expanding and reducing the pixel block of the fingerprint sample data, including:

s122: obtaining a first gray compensation value according to gray values of pixel blocks of a fingerprint part and a background part in the first sample data; the first gray compensation value is obtained through the following formula:

H _xa ＝ln(1+H _x2 -H _x1 )

wherein H is _xa Representing a first gray scale compensation value; h _x1 And H _x2 Representing a first compensation component and a second compensation component, respectively; m represents the number of pixel blocks after the size of the pixel block is enlarged; n represents the number of pixel blocks before the size of the pixel block is enlarged; h _di A gradation value representing an i-th pixel block after the size of the pixel block is enlarged; h _yi A gradation value representing an i-th pixel block before the size of the pixel block is enlarged; h _b An average gray value of the pixel block representing the background portion; sqrt represents an open root number operation;

s124: obtaining a second gray level compensation value according to gray level values of pixel blocks of the fingerprint part and the background part in the second sample data; the second gray level compensation value is obtained through the following formula:

H _xb ＝ln(1+H _x3 -H _x2 )

wherein H is _xb Representing a second gray level compensation value; h _x3 Representing a third compensation component; k represents the number of pixel blocks after the size of the pixel block is reduced; h _si A gradation value representing an i-th pixel block after the size of the pixel block is reduced;

s125: and acquiring a comprehensive gray scale compensation value by using the first gray scale compensation value and the second gray scale compensation value. The comprehensive gray compensation value is obtained through the following formula:

wherein H is _z Representing the integrated gray scale compensation value; h _p Representing the average gray value of the fingerprint portion of each fingerprint sample data.

The working principle of the technical scheme is as follows: performing pixel block expansion on fingerprint sample data for initial training to obtain fingerprint sample data with expanded pixel blocks as first sample data; wherein the pixel block has an enlarged size range of 1.5D-3D, D representing the size of the pixel block of fingerprint sample data for initial training; obtaining a first gray compensation value according to gray values of pixel blocks of a fingerprint part and a background part in the first sample data; performing pixel block reduction on fingerprint sample data for initial training to obtain fingerprint sample data with reduced pixel blocks as second sample data; wherein the size range of the pixel block is 0.1D-0.5D, and D represents the size of the pixel block of the fingerprint sample data for initial training; obtaining a second gray level compensation value according to gray level values of pixel blocks of the fingerprint part and the background part in the second sample data; performing pixel block reduction on fingerprint sample data for initial training to obtain fingerprint sample data with reduced pixel blocks as second sample data; wherein the size range of the pixel block is 0.1D-0.5D, and D represents the size of the pixel block of the fingerprint sample data for initial training; obtaining a second gray level compensation value according to gray level values of pixel blocks of the fingerprint part and the background part in the second sample data; performing pixel block reduction on fingerprint sample data for initial training to obtain fingerprint sample data with reduced pixel blocks as second sample data; wherein the size range of the pixel block is 0.1D-0.5D, and D represents the size of the pixel block of the fingerprint sample data for initial training; obtaining a second gray level compensation value according to gray level values of pixel blocks of the fingerprint part and the background part in the second sample data; and acquiring a comprehensive gray scale compensation value by using the first gray scale compensation value and the second gray scale compensation value. The working principle of the technical scheme is as follows:

The technical scheme has the effects that: the method for obtaining the comprehensive gray compensation value by changing the size of the pixel block can effectively improve the accuracy of obtaining the comprehensive gray compensation value. By performing pixel block expansion and contraction operation on fingerprint sample data, fingerprint image information under different scales can be acquired, so that the system can capture the characteristics and details of fingerprint images more comprehensively; the gray scale difference between the fingerprint part and the background part under different scales is compensated by calculating a first gray scale compensation value and a second gray scale compensation value, so that the gray scale characteristics of the fingerprint image under different scales are more accurately described; the first gray level compensation value and the second gray level compensation value are utilized to calculate and obtain a comprehensive gray level compensation value, so that the gray level characteristics of fingerprint images under different scales can be comprehensively considered, and the adaptability of the fingerprint identification system to fingerprint images of various scales is improved; by comprehensively utilizing the multi-scale information and the gray level compensation value, the fingerprint identification system has stronger robustness and accuracy when facing factors such as different scales, illumination conditions and the like, and the practical application value of the system is improved; by carrying out multi-scale processing on the initial training fingerprint sample data, more comprehensive and diversified data support can be provided for model training, and the generalization capability and accuracy of a fingerprint identification system can be improved. The first gray level compensation value formula considers the multi-scale factors and gray level differences, comprehensively considers gray level characteristics of the fingerprint part and the background part, enhances the processing capacity and accuracy of the multi-scale fingerprint image, and is beneficial to improving the robustness and accuracy of a fingerprint identification system. Meanwhile, the first compensation component and the second compensation component in the formula respectively consider the gray value difference before and after the pixel block size is enlarged, and the gray value change condition of the fingerprint part and the background part is more comprehensively described by comparing the gray values of the pixel blocks with different sizes; the m and n parameters in the formula respectively represent the number of pixel blocks after the size of the pixel block is enlarged and reduced, and the gray level compensation value can fully synthesize fingerprint image information under different scales by considering the gray level values of the pixel blocks with different scales, so that the processing capacity of the system on the multi-scale fingerprint images is improved; h in the formula _b Parameter meterThe average gray value of the pixel blocks of the background part is shown, the gray characteristics of the fingerprint part and the background part can be better distinguished by introducing the gray value of the background part into calculation, and the accuracy of the gray compensation value is enhanced. The sqrt operation in the formula is used for amplifying and balancing the gray value difference, so that the gray compensation value can reflect gray change conditions under different scales, and the sensitivity to the gray characteristics of the fingerprint image is improved. The second gray level compensation value calculation formula comprehensively considers the multi-scale factors and gray level differences, and enhances the processing capacity and accuracy of the multi-scale fingerprint image by processing the gray level values of the pixel blocks under different scales, thereby being beneficial to improving the robustness and accuracy of the fingerprint identification system. Meanwhile, the third compensation component H in the formula _x3 The gray value difference of the pixel blocks with reduced sizes is comprehensively considered, and the gray change condition of the fingerprint part and the background part is more comprehensively described by processing the gray values of the pixel blocks with reduced sizes. The k parameter in the formula represents the number of pixel blocks after the size of the pixel blocks is reduced, and the second gray compensation value can fully integrate the fingerprint image information under different scales by considering the gray values of the pixel blocks with different scales, so that the processing capacity of the system on the multi-scale fingerprint images is improved. H in the formula _si The parameter represents the gray value of the ith pixel block after the pixel block is reduced in size, and the gray value of the pixel block with reduced size is processed, so that the second gray compensation value can reflect gray change conditions under different scales, and the sensitivity to the gray characteristics of the fingerprint image is enhanced. The comprehensive gray compensation value formula comprehensively considers the gray characteristics of fingerprint sample data, globally processes the gray information of the fingerprint image, reduces the influence of a plurality of samples on the system, is beneficial to improving the processing capacity and accuracy of a fingerprint identification system on different fingerprint samples, and enhances the robustness of the system. Meanwhile, H in the formula _p The parameter represents the average gray value of the fingerprint part of each fingerprint sample data, and the average gray value of each fingerprint sample data is introduced, so that the comprehensive gray compensation value can consider gray characteristic differences among different fingerprint samples, thereby enhancing the processing capacity of the different fingerprint samples. Comprehensive synthesisGray compensation value H _z The average gray value of the fingerprint part of each fingerprint sample data is processed, and gray characteristics of the whole fingerprint image are comprehensively considered, so that the comprehensive gray compensation value can more comprehensively describe gray change conditions of the fingerprint part and the background part. By introducing the average gray value of each fingerprint sample data, the formula can better process the gray characteristics of the individual samples, reduce the influence of the individual samples on the whole system, and improve the stability and the robustness of the system.

EXAMPLE 4,

In this embodiment, gray compensation processing is performed on a fingerprint boundary pixel block of a fingerprint reference image of an entered user at a fingerprint reference image end, so as to obtain a fingerprint reference image after gray compensation processing, including:

s22: extracting a comprehensive gray level compensation value;

s23: acquiring a first gray level compensation amount corresponding to gray level compensation processing of the fingerprint reference image end through the gray level value and the comprehensive gray level compensation value of the pixel block of the fingerprint reference image; the first gray compensation amount is obtained through the following formula:

wherein H is ₀₁ Representing a first gray scale compensation amount; r represents the number of pixel blocks of the fingerprint reference image; hri represents the gray value of the ith pixel block of the fingerprint reference image; h _z Representing the integrated gray scale compensation value; c (C) ₁ A number of pixel blocks representing a fingerprint portion of the fingerprint reference image; c (C) ₂ The number of pixel blocks representing the background portion of the fingerprint reference image; lambda (lambda) ₁ Representing a first adjustment factor;

The working principle of the technical scheme is as follows: extracting gray values of pixel blocks of the fingerprint reference image of the input user; the system processes the fingerprint reference image entered into the user, divides the image into different pixel blocks, and extracts the gray value of each pixel block. These gray values reflect the brightness or gray level of the various regions in the fingerprint image; the extraction of the integrated gray scale compensation value may involve analyzing the entire fingerprint reference image to determine the gray scale characteristics and compensation requirements as a whole. This value may be calculated based on factors such as overall brightness, contrast, etc. of the image for subsequent gray scale compensation processing. By combining the gray values of the pixel blocks of the fingerprint reference image with the integrated gray compensation values, the system calculates a first gray compensation amount, which is typically used for gray adjustment for the entire image, to achieve an overall gray compensation process. And finally, the system adjusts the pixel value of the fingerprint boundary pixel block of the fingerprint reference image to be a first gray compensation amount to form the fingerprint reference image after gray compensation processing. This step corresponds to fine gray scale adjustment of local areas of the image to enhance the features of the fingerprint and improve the quality of the image.

The technical scheme has the effects that: the gray level compensation processing is carried out on the fingerprint reference image, so that the interference caused by external factors such as illumination conditions, influence and the like can be effectively reduced, and the accuracy and the robustness of fingerprint identification are improved; the gray level compensation process can help to highlight and enhance the characteristics in the fingerprint image, so that the fingerprint image has more distinction, thereby improving the reliability of fingerprint identification; the gray level compensation processing is helpful to optimize the quality of the fingerprint image by adjusting the pixel value of the fingerprint reference image, so that the image is clearer and easier to analyze and identify; the extraction of the comprehensive gray compensation value and the adjustment of the gray compensation amount can be adjusted according to different environmental conditions and requirements, so that the fingerprint identification system can be better adapted to different working scenes. By the first gray level compensationThe calculation formula can enable the calculation of gray compensation quantity to be more intelligent and personalized, and can be better suitable for the characteristics of different fingerprint images, so that the performance of the whole fingerprint identification system is improved. Meanwhile, the formula can be adjusted in a personalized way according to the characteristics of the specific fingerprint reference image. By considering the number of pixel blocks of the fingerprint part and the background part, the formula can compensate gray features of different parts to different degrees so as to adapt to the characteristics of different fingerprint images. The formula contains the comprehensive gray compensation value H _z This enables the calculation of the gray compensation amount to comprehensively consider gray features of the entire fingerprint image, not just gray values of local areas. Therefore, the gray level of the image can be more comprehensively adjusted, and the quality of the fingerprint image is improved. Adjustment coefficient lambda ₁ The gray compensation quantity calculation of the formula is adjustable, parameters can be adjusted according to actual conditions, the optimal gray compensation effect is obtained, and the gray of the fingerprint reference image can be adjusted more accurately through the personalized, comprehensive consideration and adjustable gray compensation calculation, so that the accuracy and the robustness of the fingerprint identification system on the fingerprint image are improved.

EXAMPLE 5,

In this embodiment, when a fingerprint image to be detected of a user is obtained, gray compensation processing is performed on a pixel block at a fingerprint edge of the fingerprint image to be detected at an image end to form a target fingerprint image to be detected, including:

s32: extracting a comprehensive gray level compensation value;

s33: acquiring a second gray level compensation amount corresponding to gray level compensation processing of the image end to be detected through the gray level value and the comprehensive gray level compensation value of the pixel block of the fingerprint image to be detected; the second gray level compensation amount is obtained through the following formula:

Wherein H is ₀₂ Representing a second gray level compensation amount; t represents the number of pixel blocks of the fingerprint image to be detected; hti the gray value of the ith pixel block of the fingerprint image to be detected; h _z Representing the integrated gray scale compensation value; c (C) ₃ The number of pixel blocks representing the fingerprint portion of the fingerprint image to be detected; c (C) ₄ Representing the number of pixel blocks referring to the background portion of the fingerprint image to be detected; lambda (lambda) ₂ Representing a second adjustment factor.

The working principle of the technical scheme is as follows: and carrying out pixel block segmentation on the fingerprint image to be detected, wherein each pixel block comprises a certain number of pixels. Then, the gray values of the pixel blocks are extracted. The gray value represents color information of each pixel ranging from 0 to 255, where 0 represents black and 255 represents white. This process is to acquire luminance information of each pixel in the image; the extraction of the integrated gray-scale compensation value is to consider the brightness and contrast of the whole image. The method is obtained by processing various statistical characteristics such as average gray value, maximum gray value, minimum gray value and the like of the whole image. The integrated gray-scale compensation value is intended to provide a reference for the subsequent gray-scale compensation process; and processing the gray value and the comprehensive gray compensation value of the pixel block of the fingerprint image to be detected to obtain a second gray compensation amount corresponding to gray compensation processing of the image end to be detected. This process typically involves some mathematical operations, such as addition, subtraction, multiplication, etc., to effect correction of the gray value of each pixel; and adjusting the pixel value of the fingerprint boundary pixel block of the fingerprint image to be detected to be a second gray level compensation amount.

The technical scheme has the effects that: the gray level of the edge area can be extracted by carrying out gray level compensation processing on the fingerprint edge pixel blocks of the fingerprint image to be detectedThe value is increased, so that the information of the fingerprint edge is enhanced, and the boundary of the target fingerprint image to be detected is clearer and more accurate; the second gray level compensation amount corresponding to the gray level compensation processing of the image end to be detected can be calculated by extracting the comprehensive gray level compensation value and combining the gray level value of the pixel block of the fingerprint image to be detected. By the aid of the method, local gray level distribution balance of the fingerprint image can be achieved, brightness difference of the whole image is smaller, and a better visual effect is achieved; the pixel value of the fingerprint boundary pixel block of the fingerprint image to be detected is adjusted to be the second gray level compensation amount, so that gray level difference of the edge area can be effectively corrected, quality and definition of the fingerprint image to be detected are further enhanced, and more accurate input is provided for subsequent fingerprint identification. The formula combines the pixel block gray value, the comprehensive gray compensation value and the adjustment coefficient of the fingerprint image to be detected, can individually compensate gray, adjust gray distribution, distinguish background from fingerprint, and flexibly control gray compensation effect through the adjustment coefficient, thereby improving the quality and reliability of the fingerprint image of the object to be detected. Meanwhile, according to the above formula, hti in the formula represents the gray value of each pixel block according to the gray value of the pixel block of the fingerprint image to be detected, and the average gray value can be obtained by summing all the pixel blocks and dividing by the number t of the pixel blocks. Thus, the gray scale characteristics of the whole fingerprint image can be reflected more accurately, and personalized gray scale compensation is realized; comprehensive gray level compensation value H _z The function of adjusting the gray distribution is played in the formula. By comprehensively considering the local gray scale characteristics and the whole gray scale compensation value of the fingerprint image to be detected, the proper gray scale adjustment can be carried out on different areas according to specific conditions, so that the visual effect of the whole image is improved; c in the formula ₃ And C ₄ The number of pixel blocks respectively representing the fingerprint part and the background part in the fingerprint image to be detected. By considering the influence of the background part in the calculation, the fingerprint and the background can be better distinguished, and the definition and the discernability of the fingerprint image of the object to be detected are further improved; adjustment coefficient lambda ₂ The function of adjusting gray compensation is played in the formula. The intensity and effect of gray compensation can be flexibly controlled by adjusting the value of the coefficient toAdapt to different practical application demands.

EXAMPLE 6,

In this embodiment, a fingerprint image recognition system based on deep learning, the fingerprint image recognition system includes:

and a data processing module: carrying out gray level compensation processing on fingerprint sample data in a database by using a fingerprint boundary pixel block, and carrying out model training twice by using the fingerprint sample data subjected to the gray level compensation processing and an original fingerprint sample data deep learning model to finish a trained deep learning model; specifically, the deep learning model is trained twice by combining two image pixel conditions of enlarging pixel blocks and shrinking pixel blocks of fingerprint sample data in a database and a mode of gray compensation processing of fingerprint boundary pixel blocks of the fingerprint sample data, so that the trained deep learning model is completed;

EXAMPLE 7,

In this embodiment, the data processing module includes:

and a data extraction module: extracting fingerprint sample data from a database, and utilizing 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; the deep learning model adopts a model with a convolutional neural network structure;

and a data acquisition module: performing gray level compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data for initial training through the comprehensive gray level compensation value to obtain adjusted sample data; specifically, the first gray level compensation value and the second gray level compensation value are utilized to obtain a comprehensive gray level compensation value, and the comprehensive data compensation value is utilized to carry out gray level compensation processing on the fingerprint boundary pixel block of the fingerprint sample data for initial training, so that the gray level value of the fingerprint boundary pixel block reaches the comprehensive gray level compensation value;

EXAMPLE 8,

In this embodiment, the compensation value obtaining module includes:

a first gray compensation value module: obtaining a first gray compensation value according to gray values of pixel blocks of a fingerprint part and a background part in the first sample data; the first gray compensation value is obtained through the following formula:

H _xa ＝ln(1+H _x2 -H _x1 )

the second gray level compensation value module: obtaining a second gray level compensation value according to gray level values of pixel blocks of the fingerprint part and the background part in the second sample data; the second gray level compensation value is obtained through the following formula:

H _xb ＝ln(1+H _x3 -H _x2 )

The comprehensive gray compensation value is obtained through the following formula:

The technical scheme has the effects that: the method for obtaining the comprehensive gray compensation value by changing the size of the pixel block can effectively improve the accuracy of obtaining the comprehensive gray compensation value. By performing pixel block expansion and contraction operation on fingerprint sample data, fingerprint image information under different scales can be acquired, so that the system can capture the characteristics and details of fingerprint images more comprehensively; the gray scale difference between the fingerprint part and the background part under different scales is compensated by calculating a first gray scale compensation value and a second gray scale compensation value, so that the gray scale characteristics of the fingerprint image under different scales are more accurately described; the first gray level compensation value and the second gray level compensation value are utilized to calculate and obtain a comprehensive gray level compensation value, so that the gray level characteristics of fingerprint images under different scales can be comprehensively considered, and the adaptability of the fingerprint identification system to fingerprint images of various scales is improved; by comprehensively utilizing the multi-scale information and the gray level compensation value, the fingerprint identification system has stronger robustness and accuracy when facing factors such as different scales, illumination conditions and the like, and the practical application value of the system is improved; by multi-scale processing of initially trained fingerprint sample data And more comprehensive and diversified data support can be provided for model training, and the generalization capability and accuracy of the fingerprint identification system can be improved. The first gray level compensation value formula considers the multi-scale factors and gray level differences, comprehensively considers gray level characteristics of the fingerprint part and the background part, enhances the processing capacity and accuracy of the multi-scale fingerprint image, and is beneficial to improving the robustness and accuracy of a fingerprint identification system. Meanwhile, the first compensation component and the second compensation component in the formula respectively consider the gray value difference before and after the pixel block size is enlarged, and the gray value change condition of the fingerprint part and the background part is more comprehensively described by comparing the gray values of the pixel blocks with different sizes; the m and n parameters in the formula respectively represent the number of pixel blocks after the size of the pixel block is enlarged and reduced, and the gray level compensation value can fully synthesize fingerprint image information under different scales by considering the gray level values of the pixel blocks with different scales, so that the processing capacity of the system on the multi-scale fingerprint images is improved; h in the formula _b The parameters represent the average gray value of the pixel blocks of the background part, and the gray characteristics of the fingerprint part and the background part can be better distinguished by introducing the gray value of the background part into calculation, so that the accuracy of the gray compensation value is enhanced. The sqrt operation in the formula is used for amplifying and balancing the gray value difference, so that the gray compensation value can reflect gray change conditions under different scales, and the sensitivity to the gray characteristics of the fingerprint image is improved. The second gray level compensation value calculation formula comprehensively considers the multi-scale factors and gray level differences, and enhances the processing capacity and accuracy of the multi-scale fingerprint image by processing the gray level values of the pixel blocks under different scales, thereby being beneficial to improving the robustness and accuracy of the fingerprint identification system. Meanwhile, the third compensation component H in the formula _x3 The gray value difference of the pixel blocks with reduced sizes is comprehensively considered, and the gray change condition of the fingerprint part and the background part is more comprehensively described by processing the gray values of the pixel blocks with reduced sizes. The k parameter in the formula represents the number of pixel blocks after the size of the pixel blocks is reduced, and the second gray compensation value can be fully integrated under different scales by considering the gray values of the pixel blocks with different scalesThe fingerprint image information of the system improves the processing capacity of the system on the multi-scale fingerprint image. H in the formula _si The parameter represents the gray value of the ith pixel block after the pixel block is reduced in size, and the gray value of the pixel block with reduced size is processed, so that the second gray compensation value can reflect gray change conditions under different scales, and the sensitivity to the gray characteristics of the fingerprint image is enhanced. The comprehensive gray compensation value formula comprehensively considers the gray characteristics of fingerprint sample data, globally processes the gray information of the fingerprint image, reduces the influence of a plurality of samples on the system, is beneficial to improving the processing capacity and accuracy of a fingerprint identification system on different fingerprint samples, and enhances the robustness of the system. Meanwhile, H in the formula _p The parameter represents the average gray value of the fingerprint part of each fingerprint sample data, and the average gray value of each fingerprint sample data is introduced, so that the comprehensive gray compensation value can consider gray characteristic differences among different fingerprint samples, thereby enhancing the processing capacity of the different fingerprint samples. Comprehensive gray level compensation value H _z The average gray value of the fingerprint part of each fingerprint sample data is processed, and gray characteristics of the whole fingerprint image are comprehensively considered, so that the comprehensive gray compensation value can more comprehensively describe gray change conditions of the fingerprint part and the background part. By introducing the average gray value of each fingerprint sample data, the formula can better process the gray characteristics of the individual samples, reduce the influence of the individual samples on the whole system, and improve the stability and the robustness of the system.

EXAMPLE 9,

In this embodiment, the image obtaining module includes:

the gray compensation value acquisition module: acquiring a first gray level compensation amount corresponding to gray level compensation processing of the fingerprint reference image end through the gray level value and the comprehensive gray level compensation value of the pixel block of the fingerprint reference image; the first gray compensation amount is obtained through the following formula:

The technical scheme has the effects that: the gray level compensation processing is carried out on the fingerprint reference image, so that the interference caused by external factors such as illumination conditions, influence and the like can be effectively reduced, and the accuracy and the robustness of fingerprint identification are improved; the gray level compensation process can help to highlight and enhance the characteristics in the fingerprint image, so that the fingerprint image has more distinction, thereby improving the reliability of fingerprint identification; the gray level compensation processing is helpful to optimize the quality of the fingerprint image by adjusting the pixel value of the fingerprint reference image, so that the image is clearer and easier to analyze and identify; the extraction of the comprehensive gray compensation value and the adjustment of the gray compensation amount can be adjusted according to different environmental conditions and requirements, so that the fingerprint identification system can be better adapted to different working scenes. The calculation of the gray compensation quantity can be more intelligent and personalized through the first gray compensation quantity calculation formula, and the calculation can be better adapted to the characteristics of different fingerprint images, so that the performance of the whole fingerprint identification system is improved. Meanwhile, the formula can be adjusted in a personalized way according to the characteristics of the specific fingerprint reference image. By considering the number of pixel blocks of the fingerprint part and the background part, the formula can compensate gray features of different parts to different degrees so as to adapt to the characteristics of different fingerprint images. The formula contains the comprehensive gray compensation value H _z This enables the calculation of the gray compensation amount to comprehensively consider gray features of the entire fingerprint image, not just gray values of local areas. Therefore, the gray level of the image can be more comprehensively adjusted, and the quality of the fingerprint image is improved. Adjustment coefficient lambda ₁ The gray compensation quantity calculation of the formula is adjustable, parameters can be adjusted according to actual conditions, the optimal gray compensation effect is obtained, and the gray of the fingerprint reference image can be adjusted more accurately through the personalized, comprehensive consideration and adjustable gray compensation calculation, so that the accuracy and the robustness of the fingerprint identification system on the fingerprint image are improved.

EXAMPLE 10,

In this embodiment, the image forming module includes:

the acquisition module is used for: acquiring a second gray level compensation amount corresponding to gray level compensation processing of the image end to be detected through the gray level value and the comprehensive gray level compensation value of the pixel block of the fingerprint image to be detected; the second gray level compensation amount is obtained through the following formula:

The technical scheme has the effects that: gray level compensation processing is carried out on the fingerprint edge pixel blocks of the fingerprint image to be detected, so that gray level values of edge areas can be extracted, information of the fingerprint edges is enhanced, and the boundary of the fingerprint image to be detected is clearer and more accurate; the second gray level compensation amount corresponding to the gray level compensation processing of the image end to be detected can be calculated by extracting the comprehensive gray level compensation value and combining the gray level value of the pixel block of the fingerprint image to be detected. By the aid of the method, local gray level distribution balance of the fingerprint image can be achieved, brightness difference of the whole image is smaller, and a better visual effect is achieved; the pixel value of the fingerprint boundary pixel block of the fingerprint image to be detected is adjusted to be the second gray level compensation amount, so that gray level difference of the edge area can be effectively corrected, quality and definition of the fingerprint image to be detected are further enhanced, and more accurate input is provided for subsequent fingerprint identification. The formula combines the pixel block gray value, the comprehensive gray compensation value and the adjustment coefficient of the fingerprint image to be detected, can individually compensate gray, adjust gray distribution, distinguish background from fingerprint, and flexibly control gray compensation effect through the adjustment coefficient, thereby improving the quality and reliability of the fingerprint image of the object to be detected. Meanwhile, according to the above formula, hti in the formula represents the gray value of each pixel block according to the gray value of the pixel block of the fingerprint image to be detected, and the average gray value can be obtained by summing all the pixel blocks and dividing by the number t of the pixel blocks. Thus, the gray scale characteristics of the whole fingerprint image can be reflected more accurately Thereby realizing personalized gray compensation; comprehensive gray level compensation value H _z The function of adjusting the gray distribution is played in the formula. By comprehensively considering the local gray scale characteristics and the whole gray scale compensation value of the fingerprint image to be detected, the proper gray scale adjustment can be carried out on different areas according to specific conditions, so that the visual effect of the whole image is improved; c in the formula ₃ And C ₄ The number of pixel blocks respectively representing the fingerprint part and the background part in the fingerprint image to be detected. By considering the influence of the background part in the calculation, the fingerprint and the background can be better distinguished, and the definition and the discernability of the fingerprint image of the object to be detected are further improved; adjustment coefficient lambda ₂ The function of adjusting gray compensation is played in the formula. The intensity and effect of gray level compensation can be flexibly controlled by adjusting the value of the coefficient so as to adapt to different practical application requirements.

It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.

Claims

1. The fingerprint image recognition method based on the deep learning is characterized by comprising the following steps of:

carrying out gray level compensation processing on fingerprint sample data in a database by using a fingerprint boundary pixel block, and carrying out model training twice by using the fingerprint sample data subjected to the gray level compensation processing and an original fingerprint sample data deep learning model to finish a trained deep learning model;

gray level compensation processing is carried out on a fingerprint boundary pixel block of the input fingerprint reference image of the user at a fingerprint reference image end, a fingerprint reference image after gray level compensation processing is obtained, and fingerprint characteristics of the fingerprint reference image are obtained through the deep learning model and are used as target fingerprint characteristics;

when a fingerprint image to be detected of a user is obtained, gray level compensation processing of an image end to be detected is carried out on a fingerprint edge pixel block of the fingerprint image to be detected, so that a target fingerprint image to be detected is formed;

and inputting the target fingerprint image to be detected into a deep learning model, acquiring the target fingerprint image to be detected by the deep learning model to perform fingerprint feature recognition, acquiring 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 acquire a fingerprint image recognition result.

2. The deep learning based fingerprint image recognition method of claim 1, wherein the performing gray level compensation processing of the fingerprint boundary pixel block on the fingerprint sample data in the database, and performing model training twice by using the fingerprint sample data after the gray level compensation processing and the original fingerprint sample data deep learning model, and completing the trained deep learning model comprises:

extracting fingerprint sample data from a database, and utilizing 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;

performing pixel block expansion and pixel block reduction on fingerprint sample data for initial training, and obtaining a comprehensive gray level compensation value corresponding to the fingerprint sample data for initial training by performing pixel block expansion and pixel block reduction on the fingerprint sample data;

performing gray level compensation processing on the fingerprint boundary pixel blocks of the fingerprint sample data for initial training through the comprehensive gray level compensation value to obtain adjusted sample data;

performing secondary training on the deep learning model by using the adjusted sample data to obtain a trained deep learning model;

Verifying the deep learning model by using a fingerprint verification sample to obtain the accuracy of fingerprint image feature recognition of the deep learning model;

and when the accuracy is lower than a preset accuracy threshold, optimizing the deep learning model until the verified accuracy reaches or exceeds the preset accuracy threshold.

3. The deep learning based fingerprint image recognition method of claim 2, wherein performing pixel block expansion and pixel block reduction on fingerprint sample data for initial training, and obtaining a comprehensive gray-scale compensation value corresponding to the fingerprint sample data for initial training by enlarging and reducing the pixel block of the fingerprint sample data, comprises:

performing pixel block expansion on fingerprint sample data for initial training to obtain fingerprint sample data with expanded pixel blocks as first sample data; wherein the pixel block has an enlarged size range of 1.5D-3D, D representing the size of the pixel block of fingerprint sample data for initial training;

obtaining a first gray compensation value according to gray values of pixel blocks of a fingerprint part and a background part in the first sample data;

Performing pixel block reduction on fingerprint sample data for initial training to obtain fingerprint sample data with reduced pixel blocks as second sample data; wherein the size range of the pixel block is 0.1D-0.5D, and D represents the size of the pixel block of the fingerprint sample data for initial training;

obtaining a second gray level compensation value according to gray level values of pixel blocks of the fingerprint part and the background part in the second sample data;

and acquiring a comprehensive gray scale compensation value by using the first gray scale compensation value and the second gray scale compensation value.

4. The deep learning-based fingerprint image recognition method of claim 1, wherein performing gray compensation processing on a fingerprint boundary pixel block of a fingerprint reference image of an entered user at a fingerprint reference image end to obtain a gray-compensated fingerprint reference image, comprises:

extracting gray values of pixel blocks of the registered fingerprint reference image of the user;

extracting a comprehensive gray level compensation value;

acquiring a first gray level compensation amount corresponding to gray level compensation processing of the fingerprint reference image end through the gray level value and the comprehensive gray level compensation value of the pixel block of the fingerprint reference image;

And adjusting the pixel value of the fingerprint boundary pixel block of the fingerprint reference image to be a first gray level compensation amount to form a fingerprint reference image after gray level compensation processing.

5. The deep learning-based fingerprint image recognition method according to claim 1, wherein when a fingerprint image to be detected of a user is acquired, gray compensation processing is performed on a fingerprint edge pixel block of the fingerprint image to be detected at an image end to form a target fingerprint image to be detected, including:

extracting gray values of pixel blocks of a fingerprint image to be detected;

extracting a comprehensive gray level compensation value;

acquiring a second gray level compensation amount corresponding to gray level compensation processing of the image end to be detected through the gray level value and the comprehensive gray level compensation value of the pixel block of the fingerprint image to be detected;

and adjusting the pixel value of the fingerprint boundary pixel block of the fingerprint image to be detected to be a second gray level compensation amount to form the target fingerprint image to be detected.

6. A deep learning based fingerprint image recognition system, the fingerprint image recognition system comprising:

7. The deep learning based fingerprint image recognition system of claim 6, wherein the data processing module comprises:

8. The deep learning based fingerprint image recognition system of claim 7, wherein the compensation value obtaining module comprises:

9. The deep learning based fingerprint image recognition system of claim 6, wherein the image acquisition module comprises:

10. The deep learning based fingerprint image recognition system of claim 6, wherein the image forming module comprises: