A deep learning-based fingerprint image recognition conversion method

Abstract

The application discloses a kind of based on deep learning fingerprint image recognition conversion method, it is related to fingerprint image recognition conversion technical field, including the following steps: skin fingerprint feature is extracted to fingerprint image, the fingerprint data of fingerprint image is recorded, image quality training set, feature information analysis is carried out to fuzzy fingerprint image training set, quality evaluation test after feature comparison and the feature of fuzzy fingerprint image training set and the feature of clear fingerprint image training set are analyzed and compared and the feature of fuzzy fingerprint image training set is analyzed and compared with the feature of clear fingerprint image training set.The based on deep learning fingerprint image recognition conversion method, by feature information analysis to fuzzy fingerprint image training set, can be through the analysis and comparison of big data, realize the feature information optimization of fuzzy fingerprint image training set, again in combination with the feature information in image quality training set, can quickly and effectively establish fingerprint image database, so as to be able to guarantee the efficiency of equipment to fingerprint recognition conversion.

Description

Technical Field

[0001] This invention relates to the field of fingerprint image recognition and conversion technology, specifically a fingerprint image recognition and conversion method based on deep learning. Background Technology

[0002] Biometric identification methods include facial, voice, iris, retinal, vein, and fingerprint recognition. Fingerprints are the ridges and grooves on the skin at the tips of a fingertips. Because each person's fingerprints are unique and do not easily change with age or health, fingerprint recognition has become the most widely used biometric method. With the continuous development of computer image processing and pattern recognition technologies, biometric technology is being applied more and more extensively.

[0003] Fingerprint recognition technology is one of many biometric identification technologies. Biometric identification technology refers to the use of inherent physiological or behavioral characteristics of the human body to identify an individual. Fingerprint recognition classifies and compares the fingerprints of the identified person to make a judgment. As one of the biometric identification technologies, fingerprint recognition technology has gradually matured in the new century and entered the fields of human production and life. In recent years, fingerprint recognition technology has developed rapidly and is one of the more mature identification methods among many biometric identification technologies. Moreover, with the rise of smartphones, fingerprint recognition has been widely used in the field of smartphones: unlocking phones, payment information, message confirmation, etc.

[0004] In existing technologies, fingerprint image recognition typically involves a combination of manual and mechanical methods to determine image feature points. When these feature points are blurry, the device cannot accurately identify the relevant image information. This results in existing devices failing to quickly and accurately recognize fingerprints due to unclear fingerprints, leading to poor performance. To address this issue, we propose a deep learning-based fingerprint image recognition and conversion method. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides a fingerprint image recognition and conversion method based on deep learning, which solves the problems mentioned in the background section.

[0007] (II) Technical Solution

[0008] To achieve the above objectives, the present invention provides the following technical solution: a fingerprint image recognition and conversion method based on deep learning, comprising the following steps:

[0009] S1. Extract skin fingerprint features from the fingerprint image;

[0010] S2. Record the fingerprint data of the fingerprint image. The fingerprint data includes the location information of the fingerprint image and the authenticity information of the fingerprint image. If the collected fingerprint data does not match the skin characteristics, the fingerprint image is a fake fingerprint image. Conversely, if the collected fingerprint data matches the skin characteristics, the fingerprint image is a real fingerprint image.

[0011] S3, Image Quality Training Set: This set filters and distinguishes the collected real fingerprint images based on their image quality. It divides the collected fingerprint images into a fuzzy fingerprint image training set and a clear fingerprint image training set. Information from these two sets is labeled using a labeling module and stored using an information storage module. By differentiating and filtering fake fingerprint images, the training efficiency of fingerprint images can be improved. Furthermore, by setting up the image quality training set, images in the fuzzy fingerprint image training set can be used for training, ensuring the true training effect of fingerprint images and achieving rapid training of fingerprint images, thus guaranteeing the final fingerprint image recognition and conversion efficiency.

[0012] S4. Perform feature information analysis on the fuzzy fingerprint image training set, label and remove data that do not meet the feature information analysis criteria, and train and optimize the fuzzy fingerprint image training set. Then, compare the features of the trained fuzzy fingerprint image training set with those of the clear fingerprint image training set.

[0013] S5. Quality evaluation test after feature comparison: A secondary fingerprint recognition test is performed on the data of the fuzzy fingerprint image training set. The test results are evaluated using big data through neural network analysis. By analyzing the feature information of the fuzzy fingerprint image training set, the feature information of the fuzzy fingerprint image training set can be optimized through big data analysis and comparison. Combined with the feature information in the image quality training set, the fingerprint image database can be established quickly and effectively, thereby ensuring the efficiency of the device in fingerprint recognition conversion.

[0014] S6. Analyze and compare the features of the fuzzy fingerprint image training set with the features of the clear fingerprint image training set, and extract the intersection data of the features of the fuzzy fingerprint image training set and the features of the clear fingerprint image training set. Analyze and save the extracted intersection data to establish an image training set feature library. Repeat the above steps to improve the data of the image training set feature library. By cleaning up redundant data, the memory usage of the storage device can be kept at a normal level. By repeatedly training the fingerprint information features, the data update of the image training set feature library can be guaranteed. By establishing a larger image training set, the device can quickly deal with different types of fingerprint images, thereby ensuring the device's fingerprint image recognition and conversion efficiency.

[0015] Optionally, the feature information analysis includes the color space distribution values ​​of each pixel in the image, the mean and variance of the color space distribution values ​​of each pixel in the image.

[0016] Optionally, the location information of the fingerprint image includes:

[0017] An image enhancement network is applied to the on-site fingerprint image;

[0018] The enhanced on-site fingerprint image is segmented into multiple on-site fingerprint enhancement sub-image blocks.

[0019] Optionally, the image enhancement network is implemented using TP-GAN technology. Based on the characteristics of high-quality fingerprint images, the completeness and clarity of the low-quality fingerprint images after labeling are supplemented, thereby converting them into high-quality fingerprint images and constructing a quality enhancement network.

[0020] Optionally, the numerical values ​​of the feature comparison are the color space distribution values ​​of each pixel in the image, as well as the mean and variance of the color space distribution values ​​of the image pixels.

[0021] Optionally, the fingerprint image recognition and conversion method further includes: using ResNet as the base network, learning the characteristics of a clear fingerprint image training set, introducing big data information from the Internet, and constructing a neural network.

[0022] Optionally, the labeling module is used for data labeling and classification of the fuzzy fingerprint image training set and the clear fingerprint image training set.

[0023] Optionally, the intersection data accounts for 70%-80% of the total training set data of the fuzzy fingerprint images.

[0024] Optionally, the intersection data accounts for 75%-85% of the total clear fingerprint image training set data.

[0025] Optionally, S6 further includes: after analyzing and saving the extracted intersection data, the saved intersection data needs to be screened and compared, and redundant data after comparison needs to be cleaned up in a timely manner.

[0026] (III) Beneficial Effects

[0027] This invention provides a fingerprint image recognition and conversion method based on deep learning. It has the following beneficial effects:

[0028] (1) This deep learning-based fingerprint image recognition and conversion method can optimize the feature information of the fuzzy fingerprint image training set by analyzing the feature information of the training set through big data analysis and comparison. Combined with the feature information in the image quality training set, it can quickly and effectively establish a fingerprint image database, thereby ensuring the efficiency of the device in fingerprint recognition and conversion.

[0029] (2) The fingerprint image recognition and conversion method based on deep learning saves information through the information storage module. By distinguishing and filtering fake fingerprint images, it can improve the training efficiency of fingerprints. At the same time, by setting up an image quality training set, it can train on the images in the blurry fingerprint image training set, thereby ensuring the real training effect of fingerprint images, realizing the rapid training of fingerprint images, and ensuring the recognition and conversion efficiency of the final fingerprint image.

[0030] (3) The deep learning-based fingerprint image recognition and conversion method can ensure that the memory usage of the storage device is in a normal state by cleaning up redundant data, and can ensure the data update of the image training set feature library by repeatedly training the fingerprint information features. By establishing a larger image training set, the device can quickly cope with different types of fingerprint images, thereby ensuring the device's recognition and conversion efficiency of fingerprint images. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the process structure of the present invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] Example 1:

[0034] Please see Figure 1This invention provides a technical solution: a fingerprint image recognition and conversion method based on deep learning, comprising the following steps:

[0035] S1. Extract skin fingerprint features from the fingerprint image;

[0036] S2. Record the fingerprint data of the fingerprint image. The fingerprint data includes the location information of the fingerprint image and the information of whether the fingerprint image is real or fake. If the collected fingerprint data does not match the skin characteristics, the fingerprint image is a fake fingerprint image. Conversely, if the collected fingerprint data matches the skin characteristics, the fingerprint image is a real fingerprint image.

[0037] S3, Image Quality Training Set: This set filters and distinguishes the collected real fingerprint images based on their image quality. It divides the collected fingerprint images into a fuzzy fingerprint image training set and a clear fingerprint image training set. Information from these two sets is labeled using a labeling module and stored using an information storage module. By differentiating and filtering fake fingerprint images, the training efficiency of fingerprint images can be improved. Furthermore, by setting up the image quality training set, images in the fuzzy fingerprint image training set can be used for training, ensuring the true training effect of fingerprint images and achieving rapid training of fingerprint images, thus guaranteeing the final fingerprint image recognition and conversion efficiency.

[0038] S4. Perform feature information analysis on the fuzzy fingerprint image training set, label and remove data that do not meet the feature information analysis criteria, and train and optimize the fuzzy fingerprint image training set. Then, compare the features of the trained fuzzy fingerprint image training set with those of the clear fingerprint image training set.

[0039] S5. Quality evaluation test after feature comparison: A secondary fingerprint recognition test is performed on the data of the fuzzy fingerprint image training set. The test results are evaluated using big data through neural network analysis. By analyzing the feature information of the fuzzy fingerprint image training set, the feature information of the fuzzy fingerprint image training set can be optimized through big data analysis and comparison. Combined with the feature information in the image quality training set, the fingerprint image database can be established quickly and effectively, thereby ensuring the efficiency of the device in fingerprint recognition conversion.

[0040] S6. Analyze and compare the features of the fuzzy fingerprint image training set with the features of the clear fingerprint image training set, and extract the intersection data of the features of the fuzzy fingerprint image training set and the clear fingerprint image training set. Analyze and save the extracted intersection data to establish an image training set feature library. Repeat the above steps to improve the data in the image training set feature library. When analyzing and saving the extracted intersection data, it is necessary to screen and compare the saved intersection data and clean up redundant data in a timely manner. By cleaning up redundant data, the memory usage of the storage device can be kept at a normal level. By repeatedly training the fingerprint information features, the data update of the image training set feature library can be ensured. By establishing a larger image training set, the device can quickly deal with different types of fingerprint images, thereby ensuring the device's fingerprint image recognition and conversion efficiency.

[0041] Feature information analysis includes the color space distribution values ​​of each pixel in the image, the mean and variance of the color space distribution values ​​of each pixel in the image.

[0042] The location information of the fingerprint image includes:

[0043] Image enhancement network is applied to on-site fingerprint images;

[0044] The enhanced on-site fingerprint image is segmented into multiple on-site fingerprint enhancement sub-image blocks.

[0045] The image enhancement network is implemented using TP-GAN technology. Based on the characteristics of high-quality fingerprint images, it supplements the completeness and clarity of low-quality fingerprint images that have been labeled, thereby converting them into high-quality fingerprint images and constructing a quality enhancement network.

[0046] In this embodiment, the feature comparison values ​​are the color space distribution values ​​of each pixel in the image, as well as the mean and variance of the color space distribution values ​​of the image pixels. The fingerprint image recognition and conversion method further includes: using ResNet as the base network, learning the characteristics of a clear fingerprint image training set, introducing big data information from the Internet, and constructing a neural network.

[0047] It is worth noting that the labeling module is used for data labeling and classification in both the fuzzy fingerprint image training set and the clear fingerprint image training set. The intersection data accounts for 70% of the total fuzzy fingerprint image training set data. The intersection data accounts for 75% of the total clear fingerprint image training set data.

[0048] Example 2:

[0049] Please see Figure 1 This invention provides a technical solution: a fingerprint image recognition and conversion method based on deep learning, comprising the following steps:

[0050] S1. Extract skin fingerprint features from the fingerprint image;

[0051] S2. Record the fingerprint data of the fingerprint image. The fingerprint data includes the location information of the fingerprint image and the information of whether the fingerprint image is real or fake. If the collected fingerprint data does not match the skin characteristics, the fingerprint image is a fake fingerprint image. Conversely, if the collected fingerprint data matches the skin characteristics, the fingerprint image is a real fingerprint image.

[0052] S3, Image Quality Training Set: This set filters and distinguishes the image quality of the collected real fingerprint images, dividing them into a fuzzy fingerprint image training set and a clear fingerprint image training set. The information in the fuzzy fingerprint image training set and the clear fingerprint image training set is marked by the marking module and saved by the information storage module.

[0053] S4. Perform feature information analysis on the fuzzy fingerprint image training set, label and remove data that do not meet the feature information analysis criteria, and train and optimize the fuzzy fingerprint image training set. Then, compare the features of the trained fuzzy fingerprint image training set with those of the clear fingerprint image training set.

[0054] S5. Quality evaluation test after feature comparison: The test data of the fuzzy fingerprint image training set is used to perform a secondary fingerprint recognition test, and the test results are evaluated by big data through neural network analysis.

[0055] S6. Analyze and compare the features of the fuzzy fingerprint image training set with the features of the clear fingerprint image training set, and extract the intersection data of the features of the fuzzy fingerprint image training set and the features of the clear fingerprint image training set. Analyze and save the extracted intersection data to establish an image training set feature library. Repeat the above steps to improve the data of the image training set feature library. Analyze and save the extracted intersection data. It is necessary to screen and compare the saved intersection data and clean up the redundant data in a timely manner.

[0056] Feature information analysis includes the color space distribution values ​​of each pixel in the image, the mean and variance of the color space distribution values ​​of each pixel in the image.

[0057] The location information of the fingerprint image includes:

[0058] Image enhancement network is applied to on-site fingerprint images;

[0059] The enhanced on-site fingerprint image is segmented into multiple on-site fingerprint enhancement sub-image blocks.

[0060] The image enhancement network is implemented using TP-GAN technology. Based on the characteristics of high-quality fingerprint images, it supplements the completeness and clarity of low-quality fingerprint images that have been labeled, thereby converting them into high-quality fingerprint images and constructing a quality enhancement network.

[0061] In this embodiment, the feature comparison values ​​are the color space distribution values ​​of each pixel in the image, as well as the mean and variance of the color space distribution values ​​of the image pixels. The fingerprint image recognition and conversion method further includes: using ResNet as the base network, learning the characteristics of a clear fingerprint image training set, introducing big data information from the Internet, and constructing a neural network.

[0062] It is worth noting that the labeling module is used for data labeling and classification in both the fuzzy fingerprint image training set and the clear fingerprint image training set. The intersection data accounts for 80% of the total fuzzy fingerprint image training set data. The intersection data accounts for 85% of the total clear fingerprint image training set data.

[0063] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A fingerprint image recognition and conversion method based on deep learning, characterized in that: Includes the following steps: S1. Extract skin fingerprint features from the fingerprint image; S2. Record the fingerprint data of the fingerprint image. The fingerprint data includes the location information of the fingerprint image and the authenticity information of the fingerprint image. If the collected fingerprint data does not match the skin characteristics, the fingerprint image is a fake fingerprint image. Conversely, if the collected fingerprint data matches the skin characteristics, the fingerprint image is a real fingerprint image. S3, Image Quality Training Set: This set filters and distinguishes the image quality of the collected real fingerprint images, dividing them into a fuzzy fingerprint image training set and a clear fingerprint image training set. The information in the fuzzy fingerprint image training set and the clear fingerprint image training set is marked by the marking module and saved by the information storage module. S4. Perform feature information analysis on the fuzzy fingerprint image training set, label and remove data that do not meet the feature information analysis criteria, and train and optimize the fuzzy fingerprint image training set. Then, compare the features of the trained fuzzy fingerprint image training set with those of the clear fingerprint image training set. S5. Quality evaluation test after feature comparison: The test data of the fuzzy fingerprint image training set is used to perform a secondary fingerprint recognition test, and the test results are evaluated by big data through neural network analysis. S6. Analyze and compare the features of the fuzzy fingerprint image training set with the features of the clear fingerprint image training set, extract the intersection data of the features of the fuzzy fingerprint image training set and the features of the clear fingerprint image training set, analyze and save the extracted intersection data, establish an image training set feature library, repeat the above steps to improve the data of the image training set feature library. The feature information analysis includes the color space distribution value of each pixel in the image, the color space distribution value of each pixel in the image, and the mean and variance of the color space distribution value of each pixel in the image; S6 further includes: analyzing and saving the extracted intersection data, filtering and comparing the saved intersection data, and promptly cleaning up any redundant data after comparison.

2. The fingerprint image recognition and conversion method based on deep learning according to claim 1, characterized in that: The location information of the fingerprint image includes: An image enhancement network is applied to the on-site fingerprint image; The enhanced on-site fingerprint image is segmented into multiple on-site fingerprint enhancement sub-image blocks.

3. The fingerprint image recognition and conversion method based on deep learning according to claim 2, characterized in that: The image enhancement network is implemented using TP-GAN technology. Based on the characteristics of high-quality fingerprint images, it supplements the completeness and clarity of low-quality fingerprint images that have been marked, thereby converting them into high-quality fingerprint images and constructing a quality enhancement network.

4. The fingerprint image recognition and conversion method based on deep learning according to claim 1, characterized in that: The numerical values ​​of the feature comparison are the color space distribution values ​​of each pixel in the image, as well as the mean and variance of the color space distribution values ​​of the image pixels.

5. The fingerprint image recognition and conversion method based on deep learning according to claim 1, characterized in that: The fingerprint image recognition and conversion method further includes: using ResNet as the base network, learning the characteristics of a clear fingerprint image training set, introducing big data information from the Internet, and constructing a neural network.

6. The fingerprint image recognition and conversion method based on deep learning according to claim 1, characterized in that: The labeling module is used for data labeling and classification of the fuzzy fingerprint image training set and the clear fingerprint image training set.

7. The fingerprint image recognition and conversion method based on deep learning according to claim 1, characterized in that: The intersection data accounts for 70%-80% of the total training set data of fuzzy fingerprint images.

8. The fingerprint image recognition and conversion method based on deep learning according to claim 1, characterized in that: The intersection data accounts for 75%-85% of the total clear fingerprint image training set data.

Images