Fingerprint Minutiae Feature Map Encoding and Retrieval Method Based on SimCLR

Abstract

This application relates to a fingerprint minutiae feature map encoding and retrieval method based on SimCLR. The method includes: obtaining a fingerprint minutiae visualization image from fingerprint minutiae data, training a feature map encoding model based on the SimCLR model framework with the image training set of the fingerprint minutiae visualization image to obtain a trained image encoder, then obtaining a fingerprint minutiae visualization image from the fingerprint minutiae data of the fingerprint to be retrieved, inputting it into the trained image encoder to obtain a query image encoding, comparing it with the image encodings of the fingerprint minutiae data in the database, and screening out the fingerprints in the database that may have the same relationship with the fingerprint to be retrieved. The present invention can reduce the image storage features, reduce the interference of redundant information, quickly and accurately screen out most of the images in the database that clearly do not have the same relationship with the query fingerprint image, and greatly shorten the retrieval time of the database.

Description

Technical Field

[0001] This application relates to the field of biometric recognition technology, and particularly to a fingerprint minutiae feature map encoding and retrieval method based on SimCLR. Background Art

[0002] Biometric recognition technology (Biometrics) refers to using the unique biological behavior characteristics and physiological characteristics of the human body to identify personal behavior. Different from some external characteristics, biometric forgery is more difficult, so it has great convenience and security, making it a new star in the field of identity authentication and security. Available biometric recognition technologies include fingerprints, faces, voiceprints, irises, etc. Among them, fingerprint recognition is one of the most widely used biometric recognition technologies. It has characteristics such as universality, uniqueness, persistence, easy acquisition, and difficulty in deception. Therefore, it is widely used in fields such as access control and e-commerce reconnaissance. The fingerprint recognition method mainly includes two processes: feature extraction and comparison. The currently common feature extraction method internationally is minutiae feature extraction. Its specific steps include establishing a coordinate system for the fingerprint image, usually selecting the upper left corner as the coordinate origin, and taking the x-axis direction to the right and the y-axis direction downward as the square. Then, several minutiae are selected, and the pixel coordinate values of the minutiae and the minutiae texture direction are used as feature data. Among them, the texture direction uses 1° as the smallest differentiation unit, the positive x-axis direction is 0° (to the right), the negative y-axis direction is 90° (upward), the negative x-axis direction is 180° (to the left), and the y-axis square is 270° (downward). Fingerprint comparison is to quantitatively evaluate two fingerprint images and evaluate the matching relationship between the minutiae in the two fingerprint images through a similarity index.

[0003] There is a one-to-many matching mode in fingerprint recognition problems, which is to match the input query fingerprint with all the registered fingerprints in the fingerprint database one by one until the registered fingerprint with the best similarity is found or the conclusion of no corresponding registered fingerprint is given after searching the entire fingerprint database. The identification mode is mainly applied to criminal investigation fingerprint automatic identification systems, large-scale fingerprint attendance systems, and access control systems. With the popularization of fingerprint recognition applications in society, the scale of the fingerprint database for identifying personnel identities has also increased rapidly, and the fingerprint storage capacity of resident identity cards has even reached the level of hundreds of millions of people. This will also cause the "one-by-one" traversal method of the entire database to be unable to have practical application value due to the long time for each traversal.

[0004] Therefore, the existing technology has the problem of poor adaptability. Summary of the Invention

[0005] Based on this, in view of the above technical problems, it is necessary to provide a SimCLR-based fingerprint minutiae feature map encoding and retrieval method, device, computer device, and storage medium that can shorten the traversal time of the database each time.

[0006] A SimCLR-based fingerprint minutiae feature map encoding and retrieval method, the method comprising:

[0007] Obtain a training set of fingerprint minutiae data, obtain a fingerprint minutiae visualization image based on the fingerprint minutiae data, and further obtain a corresponding image training set; the fingerprint minutiae data is three-dimensional information data of fingerprint minutiae in a collected fingerprint image; each fingerprint minutia in the fingerprint minutiae visualization image is represented as a hollow circle with a direction.

[0008] Input the image training set into a feature map encoding model based on the SimCLR model framework for contrast training, so that the image representation similarity between image pairs is as large as possible, and the image representation similarity between non-image pairs is as small as possible, to obtain a trained feature map encoding model; the image pairs are two identical fingerprint pictures or different sampling pictures of the same fingerprint.

[0009] Obtain query fingerprint minutiae data of the fingerprint to be retrieved, obtain a query fingerprint minutiae visualization image based on the query fingerprint minutiae data, and input the query fingerprint minutiae visualization image into the trained feature map encoding model to obtain a query image encoding.

[0010] Compare the query image encoding with the image encodings of the fingerprint minutiae data in the database, and screen out the fingerprints in the database that may have the same relationship as the fingerprint to be retrieved.

[0011] In one embodiment, the feature map encoding model based on the SimCLR model framework includes: an image input module, a data augmentation module, and an image encoding module;

[0012] The image input module is used to input training images according to the batch size.

[0013] The data augmentation module is used to perform data augmentation on the input training images.

[0014] The image encoding module uses the EfficientNet-B0 model to obtain image encodings.

[0015] In one embodiment, the image encoding module further includes a CLIP model.

[0016] The CLIP model is connected to the data augmentation module for initializing image encoding, and then the obtained initialized image encoding is input into the EfficientNet-B0 model for fine-tuning.

[0017] In one embodiment, it further includes: inputting the visualized image of query fingerprint minutiae into the trained feature map encoding model;

[0018] The query image encoding is obtained through the trained image encoding module.

[0019] In one embodiment, it further includes: precisely matching the fingerprint to be retrieved and the fingerprint that may have the same relationship through a fingerprint recognition algorithm.

[0020] In one embodiment, the training set includes minutiae data of fingerprints extracted from different sampled pictures of the same fingerprint and sampled pictures of different fingerprints.

[0021] A fingerprint minutiae feature map encoding retrieval device based on SimCLR, the device includes:

[0022] The minutiae visualization module is used to obtain a training set of fingerprint minutiae data, obtain a visualized image of fingerprint minutiae according to the fingerprint minutiae data, and further obtain a corresponding image training set; the fingerprint minutiae data is the ternary information data of fingerprint minutiae in the collected fingerprint image; each fingerprint minutia in the visualized image of fingerprint minutiae is represented as a hollow circle with a direction.

[0023] The feature map encoding model training module is used to input the image training set into the feature map encoding model based on the SimCLR model framework for contrast training, so that the similarity of image representations between image pairs is as large as possible, and the similarity of image representations between non-image pairs is as small as possible, to obtain a trained feature map encoding model; the image pair is two identical fingerprint pictures or different sampled pictures of the same fingerprint.

[0024] The query image encoding determination module is used to obtain query fingerprint minutiae data of the fingerprint to be retrieved, obtain a visualized image of query fingerprint minutiae according to the query fingerprint minutiae data, input the visualized image of query fingerprint minutiae into the trained feature map encoding model, and obtain a query image encoding.

[0025] The fingerprint screening module is used to compare the query image encoding with the image encoding of fingerprint minutiae data in the database, and screen out the fingerprints in the database that may have the same relationship as the fingerprint to be retrieved.

[0026] A computer device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the following steps are implemented:

[0027] Obtain a training set of fingerprint minutiae data, obtain a fingerprint minutiae visualization image based on the fingerprint minutiae data, and further obtain a corresponding image training set. The fingerprint minutiae data is ternary information data of fingerprint minutiae in a collected fingerprint image. Each fingerprint minutia in the fingerprint minutiae visualization image is represented as a hollow circle with a direction.

[0028] Input the image training set into a feature map encoding model based on the SimCLR model framework for contrast training, so that the similarity of image representations between image pairs is as large as possible, and the similarity of image representations between non-image pairs is as small as possible, to obtain a trained feature map encoding model. The image pairs are two identical fingerprint pictures or different sampling pictures of the same fingerprint.

[0029] Obtain query fingerprint minutiae data of the fingerprint to be retrieved, obtain a query fingerprint minutiae visualization image based on the query fingerprint minutiae data, and input the query fingerprint minutiae visualization image into the trained feature map encoding model to obtain a query image encoding.

[0030] Compare the query image encoding with the image encodings of fingerprint minutiae data in the database, and screen out the fingerprints in the database that may have the same relationship with the fingerprint to be retrieved.

[0031] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the following steps are implemented:

[0032] Obtain a training set of fingerprint minutiae data, obtain a fingerprint minutiae visualization image based on the fingerprint minutiae data, and further obtain a corresponding image training set. The fingerprint minutiae data is ternary information data of fingerprint minutiae in a collected fingerprint image. Each fingerprint minutia in the fingerprint minutiae visualization image is represented as a hollow circle with a direction.

[0033] Input the image training set into a feature map encoding model based on the SimCLR model framework for contrast training, so that the similarity of image representations between image pairs is as large as possible, and the similarity of image representations between non-image pairs is as small as possible, to obtain a trained feature map encoding model. The image pairs are two identical fingerprint pictures or different sampling pictures of the same fingerprint.

[0034] Obtain the query fingerprint minutiae data of the fingerprint to be retrieved, obtain the query fingerprint minutiae visualization image according to the query fingerprint minutiae data, and input the query fingerprint minutiae visualization image into the trained feature map encoding model to obtain the query image encoding;

[0035] Compare the query image encoding with the image encodings of the fingerprint minutiae data in the database, and screen out the fingerprints in the database that may have the same relationship with the fingerprint to be retrieved.

[0036] The above-mentioned fingerprint minutiae feature map encoding retrieval method, device, computer device and storage medium based on SimCLR obtain the fingerprint minutiae visualization image according to the fingerprint minutiae data, train the feature map encoding model based on the SimCLR model framework with the image training set of the fingerprint minutiae visualization image to obtain the trained image encoder, and then obtain the fingerprint minutiae visualization image according to the fingerprint minutiae data of the fingerprint to be retrieved, input the trained image encoder to obtain the query image encoding, compare it with the image encodings of the fingerprint minutiae data in the database, and screen out the fingerprints in the database that may have the same relationship with the fingerprint to be retrieved. The present invention proposes to convert the fingerprint minutiae data into a fingerprint minutiae visualization image, and then screen the fingerprints through the image feature encoding network, which can reduce the image storage features and the interference of redundant information, and obtain the image representation at a faster speed; in addition, through fast image retrieval, most of the images in the database that clearly do not have the same relationship with the query fingerprint image can be quickly and accurately screened out, and a small number of remaining images can be "one by one" identified using the fingerprint recognition algorithm, which can greatly shorten the retrieval time of the database. Brief Description of the Drawings

[0037] Figure 1 It is a schematic flowchart of the fingerprint minutiae feature map encoding retrieval method based on SimCLR in an embodiment;

[0038] Figure 2 It is a schematic diagram of the fingerprint minutiae visualization image in an embodiment;

[0039] Figure 3 It is a schematic diagram of the feature map encoding model based on the SimCLR model framework in an embodiment;

[0040] Figure 4 It is a structural block diagram of the fingerprint minutiae feature map encoding retrieval device based on SimCLR in an embodiment;

[0041] Figure 5 It is an internal structure diagram of a computer device in an embodiment. Detailed Embodiments

[0042] To make the objectives, technical solutions and advantages of this application more clear and understandable, the following further details this application in combination with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.

[0043] In one embodiment, as Figure 1 shown, a fingerprint minutiae feature map encoding and retrieval method based on SimCLR is provided, including the following steps:

[0044] Step 102, obtain a training set of fingerprint minutiae data, obtain a fingerprint minutiae visualization image based on the fingerprint minutiae data, and further obtain a corresponding image training set.

[0045] The fingerprint minutiae data is the ternary information data of the fingerprint minutiae in the collected fingerprint image; each fingerprint minutia in the fingerprint minutiae visualization image is represented as a hollow circle with a direction.

[0046] Minutiae features are the most common fingerprint features internationally. Minutiae generally adopt a ternary storage format: (x, y, θ), where x, y, and θ represent the x-axis pixel coordinate, y-axis pixel coordinate, and minutia direction respectively. The fingerprint minutiae data in the present invention includes the fingerprint ID, the total number of minutiae, the ternary information of each minutia, and commas are used to separate each element between minutiae and between minutiae.

[0047] For example: 10200_0,3,20,30,40,40,60,34,100,110,89,

[0048] 10200_1,2,23,33,43,43,63,34,

[0049] A1,2,25,44,54,66,43,29,

[0050] Explanation: 3 minutiae are extracted from the fingerprint image of 10200_0, and its triple data is: (20, 30, 40), (40, 60, 34), (100, 110, 89);

[0051] Explanation: 2 minutiae are extracted from the fingerprint image of 10200_1, and its triple data is: (23, 33, 43), (43, 63, 34);

[0052] Explanation: 2 minutiae are extracted from the fingerprint image of A1, and its triple data is: (25, 44, 54), (66, 43, 29).

[0053] For two fingerprint IDs, if the characters before the "_" symbol are the same, they are from the same finger of the same person, and the number after the "_" symbol represents the number of times of collection. In real data, the average number of minutiae points extracted from each fingerprint image is about 20 - 40.

[0054] The present invention visualizes the minutiae features according to the minutiae point data of fingerprints, and the obtained visualization image is as shown in Figure 2 Compared with the minutiae feature map on the original fingerprint acquisition image, the fingerprint minutiae visualization image of the present invention reduces the image storage features, and when encoding the image through an image encoder, it can reduce the interference of redundant information and obtain the image representation at a faster speed.

[0055] Step 104: Input the image training set into the feature map encoding model based on the SimCLR model framework for contrast training, so that the similarity of the image representations between image pairs is as large as possible, and the similarity of the image representations between non-image pairs is as small as possible, to obtain the trained feature map encoding model.

[0056] Image pairs are two identical fingerprint pictures or different sampling pictures of the same fingerprint. For example, 10200_0 and 10200_1 are fingerprints of the same finger, which are the first collection and the second collection respectively, and 10200_0 and 10200_1 are image pairs; or the original image of 10200_0 and the image obtained by data augmentation are also image pairs.

[0057] The SimCLR model is a contrastive learning framework for visual representation. The basic schematic diagram of the feature map encoding model of the present invention based on the SimCLR model framework is as shown in Figure 3 shown, including: an image input module, a data augmentation module, and an image encoding module; the image input module is used to input training images according to the batch size; the data augmentation module is used to perform data augmentation on the input training images (such as rotation, mirroring, etc.); the image encoding module uses the EfficientNet-B0 model to obtain the image encoding. The solid line represents maximizing the similarity under encoding, while the dashed line represents minimizing the similarity.

[0058] The loss of the positive sample pairs of the model, that is, the same pictures or related pictures, is formally defined as

[0059]

[0060] where τ represents the temperature coefficient, which is a hyperparameter for adjustment and is set to 0.5 in the experiments of this embodiment; sim(·) represents calculating the cosine similarity of two image representations; z is the mapping representation of the picture after passing through a series of non-linear Dense (fully connected layer)-Relu (activation layer)-Dense layers after the encoder; N is the number of positive sample pairs.

[0061] The final loss function is the arithmetic mean of the losses of all positive sample pairs in the batch, that is,

[0062]

[0063] Step 106: Obtain the query fingerprint minutiae data of the fingerprint to be retrieved, obtain the visualized image of the query fingerprint minutiae according to the query fingerprint minutiae data, and input the visualized image of the query fingerprint minutiae into the trained feature map encoding model to obtain the query image encoding.

[0064] In the fingerprint retrieval stage, the fingerprint to be queried is input into the network, and the corresponding encoding is generated by the feature map encoding model and compared with the pictures in the database, so as to output the target picture with the best similarity.

[0065] Step 108: Compare the query image encoding with the image encodings of the fingerprint minutiae data in the database, and screen out the fingerprints in the database that may have the same relationship with the fingerprint to be retrieved.

[0066] The same relationship means that two fingerprint images are from the same fingerprint. The fingerprint data is initially screened by setting a similarity threshold.

[0067] The purpose of the present invention is to quickly and accurately screen out the vast majority of images in the database that clearly do not have the same relationship with the query fingerprint image. After the retrieval process is completed, a small number of remaining images are highly similar to the query fingerprint and are used for further "one-by-one" identification using the fingerprint recognition algorithm. The advantage of introducing the image retrieval technology is that when using the fingerprint recognition algorithm for "one-by-one" identification, it is not necessary to traverse the entire database, and only a small number of remaining images need to be "one-by-one" identified, which can greatly shorten the retrieval time of the database.

[0068] In the above fingerprint minutiae feature map encoding and retrieval method based on SimCLR, a fingerprint minutiae visualization image is obtained according to the fingerprint minutiae data, and the feature map encoding model based on the SimCLR model framework is trained by the image training set of the fingerprint minutiae visualization image to obtain a trained image encoder. Then, a fingerprint minutiae visualization image is obtained according to the fingerprint minutiae data of the fingerprint to be retrieved, and the query image encoding is obtained by inputting it into the trained image encoder. The query image encoding is compared with the image encodings of the fingerprint minutiae data in the database, and the fingerprints in the database that may have the same relationship as the fingerprint to be retrieved are screened out. The present invention proposes to convert the fingerprint minutiae data into a fingerprint minutiae visualization image, and then screen the fingerprint through the image feature encoding network, which can reduce the image storage features and the interference of redundant information, and obtain the image representation at a faster speed. In addition, through fast image retrieval, most of the images in the database that clearly do not have the same relationship as the query fingerprint image can be quickly and accurately screened out, and a small number of remaining images are identified one by one using the fingerprint recognition algorithm, which can greatly shorten the retrieval time of the database.

[0069] In one embodiment, the image encoding module further includes a CLIP model; the CLIP model is connected to the data augmentation module and is used for initializing the image encoding, and then the obtained initialized image encoding is input into the EfficientNet-B0 model for fine-tuning.

[0070] The CLIP model is a kind of image-text multi-modal learning. By pre-training an image encoder and a text classifier, the training objective is to make the latent codes encoded by the picture and the text match. In this embodiment, the image encoder of the CLIP model is connected to the data augmentation layer in the SimCLR model to obtain the image encoding of CLIP, and then it is fine-tuned through the EfficientNet model to maximize the similarity of different feature map encodings of the same fingerprint. The fingerprint information of each finger is encoded, such as 001, 002, etc., and different batches of the same finger are represented by 001-1, 001-2, etc. Then these encoding information are used as retrieval labels and trained together with the fingerprint images. Compared with the pure EfficientNet model, the CLIP model is more accurate in image encoding and can incorporate more fingerprint feature information.

[0071] In one embodiment, it further includes: inputting the query fingerprint minutiae visualization image into the trained feature map encoding model; obtaining the query image encoding through the trained image encoding module.

[0072] In one embodiment, it further includes: precisely matching the fingerprint to be retrieved and the fingerprints that may have the same relationship through the fingerprint recognition algorithm.

[0073] In one embodiment, the training set includes minutiae data extracted from different sampled images of the same fingerprint and sampled images of different fingerprints.

[0074] It should be understood that although Figure 1 each step in the flowchart of Figure 1 is shown in sequence according to the indication of the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless there is a clear indication in this article, the execution of these steps has no strict order restriction, and these steps can be executed in other orders. Moreover,

[0075] In one embodiment, as Figure 4 shown, a fingerprint minutiae feature map encoding and retrieval device based on SimCLR is provided, including: a minutiae visualization module 402, a feature map encoding model training module 404, a query image encoding determination module 406, and a fingerprint screening module 408, where:

[0076] The minutiae visualization module 402 is used to obtain the training set of fingerprint minutiae data, obtain the fingerprint minutiae visualization image according to the fingerprint minutiae data, and further obtain the corresponding image training set; the fingerprint minutiae data is the ternary information data of the fingerprint minutiae in the collected fingerprint image; each fingerprint minutiae in the fingerprint minutiae visualization image is represented as a hollow circle with a direction.

[0077] The feature map encoding model training module 404 is used to input the image training set into the feature map encoding model based on the SimCLR model framework for contrast training, so that the image representation similarity between image pairs is as large as possible, and the image representation similarity between non-image pairs is as small as possible, to obtain a trained feature map encoding model; the image pair is two identical fingerprint images or different sampled images of the same fingerprint.

[0078] The query image encoding determination module 406 is used to obtain the query fingerprint minutiae data of the fingerprint to be retrieved, obtain the query fingerprint minutiae visualization image according to the query fingerprint minutiae data, and input the query fingerprint minutiae visualization image into the trained feature map encoding model to obtain the query image encoding.

[0079] The fingerprint screening module 408 is used to compare the query image encoding with the image encoding of the fingerprint minutiae data in the database, and screen out the fingerprints in the database that may have the same relationship with the fingerprint to be retrieved.

[0080] The query image encoding determination module 406 is further configured to input the query fingerprint minutiae visualization image into the trained feature map encoding model; and obtain the query image encoding through the trained image encoding module.

[0081] The fingerprint screening module 408 is further configured to perform an exact match on the fingerprints to be retrieved and the fingerprints that may have the same relationship through a fingerprint recognition algorithm.

[0082] For the specific limitations of the fingerprint minutiae feature map encoding retrieval device based on SimCLR, reference can be made to the limitations of the fingerprint minutiae feature map encoding retrieval method based on SimCLR in the above text, which will not be elaborated here. Each module in the above fingerprint minutiae feature map encoding retrieval device based on SimCLR can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0083] In one embodiment, a computer device is provided. The computer device can be a terminal, and its internal structure diagram can be as Figure 5 shown. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a fingerprint minutiae feature map encoding retrieval method based on SimCLR. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covered on the display screen, or a button, a trackball, or a touchpad set on the housing of the computer device, or an external keyboard, a touchpad, or a mouse, etc.

[0084] Those skilled in the art can understand that Figure 5 the structure shown in

[0085] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps in the above method embodiment are implemented.

[0086] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiment are implemented.

[0087] Those of ordinary skill in the art can understand that all or part of the processes in the above method embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above method embodiments. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0088] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0089] The above-described embodiments merely represent several implementation manners of the present application. Their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A fingerprint minutiae feature map encoding and retrieval method based on SimCLR, characterized in that The method includes: Obtaining a training set of fingerprint minutiae data, obtaining a fingerprint minutiae visualization image based on the fingerprint minutiae data, and further obtaining a corresponding image training set; the fingerprint minutiae data is ternary information data of fingerprint minutiae in a collected fingerprint image; each fingerprint minutia in the fingerprint minutiae visualization image is represented as a hollow circle with a direction. Inputting the image training set into a feature map encoding model based on the SimCLR model framework for contrast training, so that the similarity of image representations between image pairs is as large as possible, and the similarity of image representations between non-image pairs is as small as possible, to obtain a trained feature map encoding model; the image pairs are two identical fingerprint pictures or different sampled pictures of the same fingerprint. Obtaining query fingerprint minutiae data of a fingerprint to be retrieved, obtaining a query fingerprint minutiae visualization image based on the query fingerprint minutiae data, and inputting the query fingerprint minutiae visualization image into the trained feature map encoding model to obtain a query image encoding. Comparing the query image encoding with the image encodings of fingerprint minutiae data in a database, and screening out fingerprints in the database that may have the same relationship as the fingerprint to be retrieved. The feature map encoding model based on the SimCLR model framework includes: an image input module, a data augmentation module, and an image encoding module. The image input module is used to input training images in batches according to the batch size. The data augmentation module is used to perform data augmentation on the input training images. The image encoding module uses the EfficientNet-B0 model to obtain image encodings. The image encoding module further includes a CLIP model. The CLIP model is connected to the data augmentation module, used to initialize image encoding, and then input the obtained initialized image encoding into the EfficientNet-B0 model for fine-tuning.

2. The method according to claim 1, wherein Inputting the query fingerprint minutiae visualization image into the trained feature map encoding model to obtain a query image encoding includes: Inputting the query fingerprint minutiae visualization image into the trained feature map encoding model. Obtaining a query image encoding through the trained image encoding module.

3. The method according to claim 1, wherein After comparing the query image encoding with the image encodings of fingerprint minutiae data in the database and screening out fingerprints in the database that may have the same relationship as the fingerprint to be retrieved, it further includes: Performing precise matching on the fingerprint to be retrieved and the fingerprints that may have the same relationship through a fingerprint recognition algorithm.

4. The method according to any one of claims 1 to 3, characterized in that The training set includes fingerprint minutiae data extracted from different sampled pictures of the same fingerprint and sampled pictures of different fingerprints.

5. A fingerprint minutiae feature map encoding and retrieval device based on SimCLR, which is used to implement the fingerprint minutiae feature map encoding and retrieval method according to any one of claims 1 to 4, and is characterized in that, The device includes: A minutiae visualization module, used to obtain a training set of fingerprint minutiae data, obtain a fingerprint minutiae visualization image based on the fingerprint minutiae data, and further obtain a corresponding image training set; the fingerprint minutiae data is ternary information data of fingerprint minutiae in a collected fingerprint image; each fingerprint minutia in the fingerprint minutiae visualization image is represented as a hollow circle with a direction. The feature map encoding model training module is used to input the image training set into the feature map encoding model based on the SimCLR model framework for contrast training, so that the similarity of image representations between image pairs is as large as possible, and the similarity of image representations between non-image pairs is as small as possible, to obtain a trained feature map encoding model; the image pairs are two identical fingerprint pictures or different sampling pictures of the same fingerprint. The query image encoding determination module is used to obtain the query fingerprint minutiae data of the fingerprint to be retrieved, obtain the query fingerprint minutiae visualization image according to the query fingerprint minutiae data, input the query fingerprint minutiae visualization image into the trained feature map encoding model, and obtain the query image encoding. The fingerprint screening module is used to compare the query image encoding with the image encodings of the fingerprint minutiae data in the database, and screen out the fingerprints in the database that may have the same relationship as the fingerprint to be retrieved.

6. The device according to claim 5, characterized in that, The fingerprint screening module is further used for: Performing precise matching on the fingerprint to be retrieved and the fingerprints that may have the same relationship through a fingerprint recognition algorithm.

7. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1 to 4.

Images