Fingerprint recognition method, device, equipment and storage medium based on improved deep learning model

Abstract

The present application discloses a fingerprint recognition method, device, equipment and storage medium based on an improved deep learning model, which relates to the field of fingerprint recognition technology, including: obtaining fingerprint information, generating an image according to the fingerprint information; inputting the image into the improved deep learning model, extracting features from the image through the deep separable convolution layer and the partial attention module in the improved deep learning model, and obtaining feature maps of different scales; preprocessing the feature maps of different scales through the feature pyramid network in the improved deep learning model to obtain multi-scale feature maps; performing convolution processing on the multi-scale feature maps through the head network in the improved deep learning model to generate a prediction tensor; and performing confidence judgment according to the prediction tensor to obtain a fingerprint recognition result. Based on the improved deep learning model, the accuracy and efficiency of fingerprint recognition are improved.

Description

Technical Field

[0001] The present application relates to the field of fingerprint recognition technology, and in particular to a fingerprint recognition method, device, equipment and storage medium based on an improved deep learning model. Background Art

[0002] As an important biometric identification technology, fingerprint recognition has been widely used in many fields due to its uniqueness, stability and convenience, covering key areas such as access control systems, mobile device unlocking, criminal investigation identification and financial security. However, with the increasing complexity and diversity of application scenarios, the existing common fingerprint recognition technology has gradually exposed many shortcomings in terms of accuracy. For example, the results of optical fingerprint recognition are easily affected by the external environment, resulting in low accuracy, the inability to unlock the phone, and a poor user experience. Summary of the invention

[0003] The main purpose of this application is to provide a fingerprint recognition method, device, equipment and storage medium based on an improved deep learning model, aiming to solve the technical problem of low accuracy of existing common fingerprint recognition technology.

[0004] To achieve the above objectives, the present application proposes a fingerprint recognition method based on an improved deep learning model, wherein the improved deep learning model includes: a backbone network, a feature pyramid network, and a head network, wherein the backbone network includes a deep separable convolutional layer and a partial attention module, and the fingerprint recognition method based on the improved deep learning model includes: Acquire fingerprint information, and generate an image according to the fingerprint information; Inputting the image into the improved deep learning model, extracting features of the image through the depth-separable convolutional layer and a partial attention module in the improved deep learning model, and obtaining feature maps of different scales; Preprocessing the feature maps of different scales through a feature pyramid network in the improved deep learning model to obtain a multi-scale feature map; Performing convolution processing on the multi-scale feature map through the head network in the improved deep learning model to generate a prediction tensor; A confidence judgment is performed based on the predicted tensor to obtain a fingerprint recognition result.

[0005] In one embodiment, the step of extracting features from the image through the depth-separable convolutional layer and the partial attention module in the improved deep learning model to obtain feature maps of different scales includes: Based on the deep separable convolution layer in the improved deep learning model, a corresponding deep convolution kernel is prepared according to each channel in the image, and each channel in the image is convolved with the corresponding deep convolution kernel channel by channel, so as to obtain output features; Preparing a convolution kernel of a preset size through the depth-separable convolution layer in the improved deep learning model, performing point-by-point convolution through the convolution kernel of the preset size, and combining the output features; The combined output features are divided, connected and fused through the partial attention module, so as to output feature maps of different scales.

[0006] In one embodiment, the step of preprocessing the feature maps of different scales through the feature pyramid network in the improved deep learning model to obtain a multi-scale feature map includes: Sampling the feature maps of different scales through a feature pyramid network in the improved deep learning model to perform feature fusion on the feature maps of different scales; The feature maps of different scales after sampling are spliced ​​so that the spliced ​​feature maps contain information of multiple scales, thereby obtaining a multi-scale feature map.

[0007] In one embodiment, the step of performing convolution processing on the multi-scale feature map through the head network in the improved deep learning model to generate a prediction tensor includes: Performing channel adjustment on the multi-scale feature map through the head network in the improved deep learning model so that the number of channels of the multi-scale feature map meets a preset value; A prediction tensor is generated according to the channel-adjusted multi-scale feature map, where the prediction tensor includes at least bounding box coordinates, object confidence, and category confidence.

[0008] In one embodiment, the step of performing confidence judgment according to the prediction tensor to obtain a fingerprint recognition result includes: Filter overlapping bounding boxes according to the bounding box coordinates, the multi-scale feature map comprising a plurality of bounding boxes; Based on the non-maximum suppression technology, confidence judgment is performed according to the predicted tensor to further filter the overlapping bounding boxes; Extract the first feature information in the filtered bounding box; Obtaining a first similarity by comparing the first feature information with preset feature information; If the first similarity is higher than a third preset threshold, it is determined that the fingerprint match is successful.

[0009] In one embodiment, the step of performing confidence judgment based on the prediction tensor and further filtering overlapping bounding boxes based on the non-maximum suppression technology includes: Based on the non-maximum suppression technology, judging whether to filter the corresponding bounding box according to the object confidence and the category confidence; If the object confidence is lower than a first preset threshold, filtering the corresponding bounding box; If the category confidence is lower than a second preset threshold, the corresponding bounding box is filtered.

[0010] In one embodiment, after the step of obtaining the first similarity by comparing the first feature information with the preset feature information, the method further includes: If the first similarity is not higher than a third preset threshold, it is determined that the fingerprint matching fails; Get the fingerprint module information of the terminal; Obtaining second feature information according to the fingerprint module information; Obtaining a second similarity by comparing the second feature information with the preset feature information; If the second similarity is higher than a fourth preset threshold, it is determined that the fingerprint match is successful.

[0011] In addition, to achieve the above objectives, the present application also proposes a fingerprint recognition device based on an improved deep learning model, the fingerprint recognition device based on the improved deep learning model includes: An acquisition module, used to acquire fingerprint information and generate an image according to the fingerprint information; A backbone network module, used to input the image into the improved deep learning model, extract features of the image through the depth-separable convolutional layer and the partial attention module in the improved deep learning model, and obtain feature maps of different scales; A feature pyramid network module, used for preprocessing the feature maps of different scales through the feature pyramid network in the improved deep learning model to obtain a multi-scale feature map; A head network module, used for performing convolution processing on the multi-scale feature map through the head network in the improved deep learning model to generate a prediction tensor; The recognition module is used to make a confidence judgment based on the predicted tensor to obtain a fingerprint recognition result.

[0012] In addition, to achieve the above-mentioned objectives, the present application also proposes a fingerprint recognition device based on an improved deep learning model, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the fingerprint recognition method based on the improved deep learning model as described above.

[0013] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by the processor, the steps of the fingerprint recognition method based on the improved deep learning model as described above are implemented.

[0014] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the fingerprint recognition method based on the improved deep learning model as described above.

[0015] One or more technical solutions proposed in this application obtain fingerprint information and generate an image based on the fingerprint information; input the image into the improved deep learning model, extract features from the image through the deep separable convolution layer and the partial attention module in the improved deep learning model, and obtain feature maps of different scales; pre-process the feature maps of different scales through the feature pyramid network in the improved deep learning model to obtain multi-scale feature maps; perform convolution processing on the multi-scale feature maps through the head network in the improved deep learning model to generate a prediction tensor; perform confidence judgment based on the prediction tensor to obtain a fingerprint recognition result. Based on the improved deep learning model, the accuracy and efficiency of fingerprint recognition are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0018] Figure 1 A flowchart of a fingerprint recognition method based on an improved deep learning model according to the present invention; Figure 2 A standard YOLOv8 model diagram provided in Embodiment 1 of the fingerprint recognition method based on an improved deep learning model of the present application; Figure 3 A flowchart diagram of the second embodiment of the fingerprint recognition method based on the improved deep learning model of the present application is provided; Figure 4A flowchart diagram of an implementation method of the second embodiment of the fingerprint recognition method based on the improved deep learning model of the present application is provided; Figure 5 This is a schematic diagram of the module structure of a fingerprint recognition device based on an improved deep learning model according to an embodiment of the present application; Figure 6 This is a schematic diagram of the device structure of the hardware operating environment involved in the fingerprint recognition method based on the improved deep learning model in the embodiment of the present application.

[0019] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0020] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0021] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0022] The main solutions of the embodiments of this application are: As application scenarios become increasingly complex and diverse, existing common fingerprint recognition technologies have gradually exposed many shortcomings in terms of accuracy. For example, the results of optical fingerprint recognition are easily seriously affected by the external environment, resulting in low accuracy, inability to unlock the phone, and a poor user experience.

[0023] The present application provides a solution to improve the deep learning model through a deep separable convolutional layer and a partial attention module, and performs fingerprint recognition based on the improved deep learning model, thereby improving the accuracy and efficiency of fingerprint recognition.

[0024] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a fingerprint recognition device based on an improved deep learning model, etc. The following takes a fingerprint recognition device based on an improved deep learning model as an example to illustrate this embodiment and the following embodiments.

[0025] Based on this, the present application embodiment provides a fingerprint recognition method based on an improved deep learning model, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the fingerprint recognition method based on the improved deep learning model of the present application.

[0026] In this embodiment, the fingerprint recognition method based on the improved deep learning model includes steps S10 to S50: Step S10: Acquire fingerprint information, and generate an image according to the fingerprint information.

[0027] It should be noted that fingerprint information refers to the fingerprint information on the user's finger that needs to be identified by the terminal. In addition, the terminal generally refers to a device or system that can receive, process or send information. It can be an electronic device such as a computer, mobile phone, tablet computer, or a dedicated hardware device or software interface.

[0028] It is understandable that the fingerprint information on the user's finger will be identified and processed to generate an image of the fingerprint, which will be provided to the improved deep learning model for fingerprint identification in subsequent steps.

[0029] Step S20: inputting the image into the improved deep learning model, performing feature extraction on the image through the deep separable convolutional layer and the partial attention module in the improved deep learning model, and obtaining feature maps of different scales.

[0030] It should be noted that the improved deep learning model includes: a backbone network, a feature pyramid network and a head network, wherein the backbone network includes a deep separable convolutional layer and a partial attention module.

[0031] In addition, the deep learning model refers to the YOLO (You Only Look Once) model, which is a popular deep learning target detection algorithm known for its fast speed and high accuracy. Its advantages are that it is extremely fast, can process 45 frames per second, can also understand general object representations, and can be applied to many practical scenarios, such as monitoring, drones, autonomous driving, etc.

[0032] In addition, the depthwise separable convolutional layer can perform two convolution operations: channel-by-channel convolution and point-by-point convolution. This convolution operation significantly reduces the number of parameters and computational complexity of the model while maintaining feature extraction capabilities.

[0033] In addition, the Pyramid Split Attention (PSA) module is an attention mechanism module used in the field of computer vision. It is designed to improve the performance of image classification and object detection tasks. The core idea of ​​the PSA module is to divide the input feature map into multiple groups in the channel dimension, and each group is convolved with convolution kernels of different sizes to extract feature information of different scales. This design enables the PSA module to capture multi-scale contextual information, thereby enhancing the feature expression ability of the model.

[0034] In addition, please refer to Figure 2 , Figure 2It is a standard YOLOv8 model, in which the CSPLayer_2Conv module not only applies the cross-stage partial network (CSP) method, but also introduces multiple gradient flows. This design enables the network to better learn features and improve the performance of the model. The backbone network (Backbone) in the standard YOLO model performs convolution operations on each channel of the input feature map at the same time, and the number of channels of the output feature map is determined by the number of convolution kernels; in this application, the original convolution network is replaced with a deep separable convolution in the Backbone to extract features, and a partial self-attention module is added after the partial convolution to enhance the feature extraction ability of the network, and finally outputs feature maps of different scales. Specifically, refer to Figure 2 ,This application improves the feature extraction efficiency by replacing the CSPLayer_2Conv module with a depthwise separable convolution module, and adds a PSA module after some of the convolutions to enhance the feature extraction capability of the network.

[0035] Specifically, this step is based on the deep separable convolution layer in the improved deep learning model, prepares a corresponding deep convolution kernel according to each channel in the image, and makes each channel in the image convolve channel by channel with the corresponding deep convolution kernel, so as to obtain output features; prepares a convolution kernel of a preset size through the deep separable convolution layer in the improved deep learning model, performs point-by-point convolution through the convolution kernel of the preset size, and combines the output features; divides, connects and fuses the combined output features through the partial attention module, so as to output feature maps of different scales.

[0036] Specifically, for each channel of the image, there is a corresponding deep convolution kernel. Assuming that the number of channels of the input image is X, X deep convolution kernels need to be prepared. The size of each deep convolution kernel is k×k, where k is usually an odd number, so that each channel in the image is convolved with the corresponding deep convolution kernel channel by channel to obtain the output features. It can be understood that one convolution kernel is only responsible for one channel to reduce the amount of calculation and the number of parameters; while point-by-point convolution uses 1×1 convolution kernels, the number of which is the number of output channels Y. The role of these 1×1 convolution kernels is to linearly combine the output features of different channels at each position of the deep convolution output, and can also adjust the number of channels of the feature map and fuse the information of different channels, ultimately achieving a significant reduction in the computational complexity of the model while maintaining or even improving the performance of the model; the PSA module consists of a multi-head self-attention module (Multi-Head Self-Attention, MHSA) and a feedforward network (Feed Forward Neural Network). Adding a PSA module after partial convolution can evenly divide the cross-channel features into two parts after point-by-point convolution, feed one part into the PSA module composed of MHSA and FFN, and then connect and fuse the two parts through 1×1 convolution to capture the global context information, enhance the model's ability to represent features, and finally output feature maps of different scales.

[0037] In addition, feature maps of different scales usually refer to the fact that in deep learning networks, due to the different network layers, the information granularity and range captured by feature maps at each level are also different. Shallow feature maps (low-level feature maps) usually have higher resolution and can capture more detailed information, but lack semantic information; while deep feature maps (high-level feature maps) have lower resolution but are rich in semantic information and can capture more abstract concepts.

[0038] Step S30: preprocessing the feature maps of different scales through the feature pyramid network in the improved deep learning model to obtain a multi-scale feature map.

[0039] It should be noted that Feature Pyramid Networks (FPN) is a deep learning network structure for target detection. It achieves the fusion and utilization of features of different scales by constructing a multi-scale feature pyramid, thereby improving detection accuracy. The core idea of ​​FPN is to fuse feature information of different scales to obtain a feature pyramid containing rich information. In convolutional neural networks, as the convolution layer deepens, the size of the feature map gradually decreases, while the semantic information gradually increases. FPN uses this feature to combine high-level semantic information with low-level detail information through upsampling and lateral connections to form a feature pyramid containing multiple scales.

[0040] It can be understood that this step mainly samples the feature maps of different scales through the feature pyramid network in the improved deep learning model to perform feature fusion on the feature maps of different scales; splices the sampled feature maps of different scales so that the spliced ​​feature maps contain information of multiple scales, thereby obtaining a multi-scale feature map.

[0041] Specifically, the structure of FPN can be divided into three parts: a bottom-up path, a top-down path, and a lateral connection. In the bottom-up path, feature maps of different scales are processed by convolutional layers and pooling layers, and are usually reduced in size by an integer multiple of 2, i.e., downsampling. This downsampling process reduces the spatial resolution of the feature map, but can capture a wider range of contextual information while reducing the amount of computation. Upsampling is performed in the top-down path. By upsampling the feature map, the upsampled feature map has the same size as the feature map of the next layer, thereby achieving feature fusion. Splicing is a lateral connection, and its essence is pixel addition. Therefore, before connecting, the high-level feature map must be upsampled and then horizontally spliced ​​with the features of the previous layer. The previous layer must undergo a 1x1 convolution before it can be spliced ​​with the upsampled layer. This step will be iterative. In this way, feature maps of different scales are fused at multiple levels to form a feature map containing multi-scale information (i.e., a multi-scale feature map).

[0042] It can be understood that FPN realizes the fusion and utilization of multi-scale features by constructing a feature pyramid, thereby improving the accuracy and efficiency of target detection.

[0043] Step S40: performing convolution processing on the multi-scale feature map through the head network in the improved deep learning model to generate a prediction tensor.

[0044] It should be noted that the head network usually refers to the network structure used for object detection or segmentation tasks. In addition, the prediction tensor is the output result obtained by the model after processing the input data. For example, in an image recognition task, the output tensor of the model may contain the probability distribution of each category, indicating the model's prediction and confidence that the input image belongs to each category.

[0045] In addition, please refer to Figure 2 In this application, the Backbone part and the Head part in the standard YOLOv8 model are mainly used. The Head part is the head network. The feature pyramid described above is obtained through the Head part in the standard YOLOv8 model.

[0046] Specifically, the multi-scale feature map is channel-adjusted through the head network in the improved deep learning model so that the number of channels of the multi-scale feature map meets the preset value; a prediction tensor is generated according to the multi-scale feature map after channel adjustment, and the prediction tensor includes at least bounding box coordinates, object confidence and category confidence.

[0047] In addition, so that the number of channels of the multi-scale feature map meets the preset value refers to adjusting the number of channels so that subsequent prediction tensor generation operations can be performed. The preset value of the number of channels after the 1x1 convolution is defined by the user. For example, the preset value of the number of channels is set to 32, and the number of channels of the feature map is changed to 32 by performing a 1x1 convolution on the multi-scale feature map, so that the number of channels meets the requirements of subsequent operations.

[0048] Step S50: Perform confidence judgment based on the predicted tensor to obtain a fingerprint recognition result.

[0049] It should be noted that the multi-scale feature map contains several bounding boxes. The confidence judgment is mainly to retain the bounding boxes with high confidence and filter the bounding boxes with low confidence, so as to obtain the fingerprint recognition result according to the filtered bounding boxes.

[0050] This embodiment provides a fingerprint recognition method based on an improved deep learning model, which obtains fingerprint information and generates an image according to the fingerprint information; inputs the image into the improved deep learning model, extracts features from the image through the deep separable convolution layer and the partial attention module in the improved deep learning model, and obtains feature maps of different scales; preprocesses the feature maps of different scales through the feature pyramid network in the improved deep learning model to obtain multi-scale feature maps; performs convolution processing on the multi-scale feature maps through the head network in the improved deep learning model to generate a prediction tensor; performs confidence judgment based on the prediction tensor to obtain a fingerprint recognition result. Based on the deep separable convolution layer and the partial attention module in the improved deep learning model, while maintaining the feature extraction capability, the number of parameters and the amount of calculation of the model are significantly reduced, and the fusion and utilization of multi-scale features are realized through the feature pyramid network, ultimately achieving the technical effect of improving the accuracy and efficiency of fingerprint recognition.

[0051] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be repeated in the following. Figure 3 , step S50 includes steps S501 to S505: Step S501: filtering overlapping bounding boxes according to the bounding box coordinates.

[0052] It should be noted that the multi-scale feature map contains several bounding boxes. First, the intersection over union (IoU) can be calculated according to the coordinates of the bounding boxes. IoU is an indicator to measure the degree of overlap between two boxes, and those bounding boxes with high overlap are removed. This can reduce the amount of calculation for further filtering overlapping bounding boxes based on non-maximum suppression technology in the following steps, thereby improving the efficiency of further filtering overlapping bounding boxes based on non-maximum suppression (NMS) technology.

[0053] Step S502: Based on the non-maximum suppression technology, confidence judgment is performed according to the prediction tensor to further filter overlapping bounding boxes.

[0054] It should be noted that non-maximum suppression technology is a post-processing technology widely used in computer vision tasks such as target detection and image segmentation. It is used to remove redundant prediction boxes and retain the prediction boxes that are most likely to represent the target object, thereby improving the accuracy and reliability of the detection results.

[0055] Specifically, based on the non-maximum suppression technology, it is determined whether to filter the corresponding bounding box according to the object confidence and the category confidence; if the object confidence is lower than a first preset threshold, the corresponding bounding box is filtered; if the category confidence is lower than a second preset threshold, the corresponding bounding box is filtered.

[0056] It can be understood that the first preset threshold is the threshold for judging the object confidence level. Object confidence refers to the model's confidence level that the target object is contained in the bounding box and the accuracy of the bounding box to the target object's position; and the second preset threshold is the threshold for judging the category confidence level. Category confidence refers to the model's credibility in the prediction result of a certain category. It expresses the possibility that the model believes that the input data belongs to a specific category in the form of a probability value.

[0057] Specifically, first, all prediction boxes are sorted in descending order according to their corresponding confidence scores. The confidence score indicates the model's confidence that the prediction box contains the target object. It is usually a value between 0 and 1. The closer it is to 1, the higher the confidence. By setting the first preset threshold and the second preset threshold, the bounding boxes whose object confidence is lower than the first preset threshold are first filtered out, and then the bounding boxes whose category confidence is lower than the second preset threshold are filtered out.

[0058] Step S503: extracting the first feature information in the filtered boundary box.

[0059] It should be noted that the feature information refers to the fingerprint detail feature points contained in the bounding box (such as endpoints, bifurcation points, etc. These feature points are key elements of fingerprint recognition, and their number, position, direction and other information can uniquely identify a fingerprint) and texture features (such as the direction, density, curvature of the grain lines, etc., which can further describe the characteristics of the fingerprint. These texture features help to construct the feature vector of the fingerprint). The first feature information refers to the feature information in the filtered bounding box obtained by processing the deep learning model improved by this application.

[0060] Step S504: Obtain a first similarity by comparing the first feature information with preset feature information.

[0061] It should be noted that the preset characteristic information refers to the characteristic information of the user fingerprint entered into the terminal. By comparing the first characteristic information with the preset characteristic information, the first similarity can be calculated.

[0062] Step S505: If the first similarity is higher than a third preset threshold, it is determined that the fingerprint match is successful.

[0063] It should be noted that the third preset threshold refers to the similarity threshold set by the user after comparing the first feature information and the preset feature information. If the first similarity is higher than the third preset threshold, the fingerprint match is determined to be successful, that is, the fingerprint recognition result is determined to be a match.

[0064] In a feasible implementation, reference may be made to Figure 4 , after step S504, steps A51 to A55 may also be included: Step A51: If the first similarity is not higher than a third preset threshold, it is determined that the fingerprint matching fails.

[0065] It can be understood that if the first similarity is not higher than the third preset threshold, it is determined that the fingerprint matching fails, that is, it is determined that the fingerprint recognition result is not matched.

[0066] Step A52: Obtain the fingerprint module information of the terminal.

[0067] It should be noted that, in order to improve the accuracy of fingerprint recognition, the present application also combines the fingerprint module of the terminal for fingerprint recognition, thereby obtaining the fingerprint module information of the terminal, that is, the feature map obtained by the fingerprint module of the terminal.

[0068] Step A53: Obtain second feature information according to the fingerprint module information.

[0069] It should be noted that the second feature information refers to the feature information in the filtered boundary frame obtained through processing by the fingerprint module.

[0070] Step A54: Obtain a second similarity by comparing the second feature information with the preset feature information.

[0071] It should be noted that the second similarity can be calculated by comparing the second feature information with the preset feature information.

[0072] Step A55: If the second similarity is higher than a fourth preset threshold, it is determined that the fingerprint match is successful.

[0073] It should be noted that the fourth preset threshold refers to a similarity threshold set by the user after comparing the second feature information and the preset feature information. It can be understood that the fourth preset threshold may be the same as or different from the third preset threshold.

[0074] In addition, if the second similarity is higher than the fourth preset threshold, it is determined that the fingerprint match is successful, that is, it is determined that the fingerprint identification result is matched.

[0075] This embodiment filters overlapping bounding boxes according to the bounding box coordinates; based on the non-maximum suppression technology, the confidence judgment is made according to the predicted tensor, and the overlapping bounding boxes are further filtered; the first feature information in the filtered bounding box is extracted; the first similarity is obtained by comparing the first feature information with the preset feature information; if the first similarity is higher than the third preset threshold, it is determined that the fingerprint match is successful. The confidence judgment is made according to the predicted tensor, and the fingerprint recognition result is determined by the feature information, which improves the accuracy of fingerprint recognition.

[0076] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the fingerprint recognition method based on the improved deep learning model of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0077] This application also provides a fingerprint recognition device based on an improved deep learning model, please refer to Figure 5 , the fingerprint recognition device based on the improved deep learning model includes: The acquisition module 10 is used to acquire fingerprint information and generate an image according to the fingerprint information.

[0078] The backbone network module 20 is used to input the image into the improved deep learning model, extract features of the image through the deep separable convolutional layer and the partial attention module in the improved deep learning model, and obtain feature maps of different scales.

[0079] The feature pyramid network module 30 is used to preprocess the feature maps of different scales through the feature pyramid network in the improved deep learning model to obtain multi-scale feature maps.

[0080] The head network module 40 is used to perform convolution processing on the multi-scale feature map through the head network in the improved deep learning model to generate a prediction tensor.

[0081] The recognition module 50 is used to perform confidence judgment based on the prediction tensor to obtain a fingerprint recognition result.

[0082] The fingerprint recognition device based on the improved deep learning model provided by the present application adopts the fingerprint recognition method based on the improved deep learning model in the above-mentioned embodiment, which can solve the technical problem of low accuracy of the existing common fingerprint recognition technology. Compared with the prior art, the beneficial effects of the fingerprint recognition device based on the improved deep learning model provided by the present application are the same as the beneficial effects of the fingerprint recognition method based on the improved deep learning model provided by the above-mentioned embodiment, and the other technical features of the fingerprint recognition device based on the improved deep learning model are the same as the features disclosed in the above-mentioned embodiment method, which will not be repeated here.

[0083] In one embodiment, the backbone network module 20 is also used to prepare a corresponding deep convolution kernel according to each channel in the image based on the deep separable convolution layer in the improved deep learning model, so that each channel in the image is convolved channel by channel with the corresponding deep convolution kernel, thereby obtaining output features; prepare a convolution kernel of a preset size through the deep separable convolution layer in the improved deep learning model, perform point-by-point convolution through the convolution kernel of the preset size, and combine the output features; divide, connect and fuse the combined output features through the partial attention module, thereby outputting feature maps of different scales.

[0084] In one embodiment, the feature pyramid network module 30 is also used to sample the feature maps of different scales through the feature pyramid network in the improved deep learning model to perform feature fusion on the feature maps of different scales; and to splice the sampled feature maps of different scales so that the spliced ​​feature maps contain information of multiple scales, thereby obtaining a multi-scale feature map.

[0085] In one embodiment, the head network module 40 is further used to perform channel adjustment on the multi-scale feature map through the head network in the improved deep learning model so that the number of channels of the multi-scale feature map meets a preset value; and generate a prediction tensor according to the multi-scale feature map after channel adjustment, wherein the prediction tensor includes at least bounding box coordinates, object confidence, and category confidence.

[0086] In one embodiment, the recognition module 50 is further used to filter overlapping bounding boxes according to the bounding box coordinates, and the multi-scale feature map includes a plurality of bounding boxes; based on the non-maximum suppression technology, confidence judgment is performed according to the prediction tensor, and the overlapping bounding boxes are further filtered; the first feature information in the filtered bounding box is extracted; by comparing the first feature information with the preset feature information, a first similarity is obtained; if the first similarity is higher than a third preset threshold, it is determined that the fingerprint match is successful.

[0087] In one embodiment, the recognition module 50 is also used to determine whether to filter the corresponding bounding box based on the non-maximum suppression technology according to the object confidence and the category confidence; if the object confidence is lower than a first preset threshold, the corresponding bounding box is filtered; if the category confidence is lower than a second preset threshold, the corresponding bounding box is filtered.

[0088] In one embodiment, the identification module 50 is also used to determine that the fingerprint matching fails if the first similarity is not higher than a third preset threshold; obtain the fingerprint module information of the terminal; obtain second feature information based on the fingerprint module information; obtain the second similarity by comparing the second feature information with the preset feature information; if the second similarity is higher than a fourth preset threshold, determine that the fingerprint matching is successful.

[0089] The present application provides a fingerprint recognition device based on an improved deep learning model, and the fingerprint recognition device based on the improved deep learning model includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the fingerprint recognition method based on the improved deep learning model in the above-mentioned embodiment 1.

[0090] Reference below Figure 6 , which shows a schematic diagram of the structure of a fingerprint recognition device based on an improved deep learning model suitable for implementing the embodiment of the present application. The fingerprint recognition device based on the improved deep learning model in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 6 The fingerprint recognition device based on the improved deep learning model shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0091] like Figure 6As shown, the fingerprint recognition device based on the improved deep learning model may include a processing device 1001 (such as a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM: Read Only Memory) 1002 or the program loaded from the storage device 1003 to the random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the fingerprint recognition device based on the improved deep learning model are also stored. The processing device 1001, ROM1002 and RAM1004 are connected to each other through a bus 1005. The input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 can allow the fingerprint recognition device based on the improved deep learning model to communicate with other devices wirelessly or by wire to exchange data. Although the figure shows a fingerprint recognition device based on an improved deep learning model with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have instead.

[0092] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0093] The fingerprint recognition device based on the improved deep learning model provided by the present application adopts the fingerprint recognition method based on the improved deep learning model in the above-mentioned embodiment, which can solve the technical problem of low accuracy of the existing common fingerprint recognition technology. Compared with the prior art, the beneficial effects of the fingerprint recognition device based on the improved deep learning model provided by the present application are the same as the beneficial effects of the fingerprint recognition method based on the improved deep learning model provided by the above-mentioned embodiment, and the other technical features of the fingerprint recognition device based on the improved deep learning model are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0094] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0095] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0096] The present application provides a computer-readable storage medium having computer-readable program instructions (i.e., computer programs) stored thereon, and the computer-readable program instructions are used to execute the fingerprint recognition method based on the improved deep learning model in the above-mentioned embodiment.

[0097] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM: Random Access Memory), a read-only memory (ROM: Read Only Memory), an erasable programmable read-only memory (EPROM: Erasable Programmable Read Only Memory or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM: CD-Read Only Memory), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency: Radio Frequency), etc., or any suitable combination of the above.

[0098] The above-mentioned computer-readable storage medium may be included in the fingerprint recognition device based on the improved deep learning model; or it may exist independently without being assembled into the fingerprint recognition device based on the improved deep learning model.

[0099] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by a fingerprint recognition device based on an improved deep learning model, the fingerprint recognition device based on the improved deep learning model: obtains fingerprint information and generates an image according to the fingerprint information; inputs the image into the improved deep learning model, and extracts features of the image through the deep separable convolution layer and the partial attention module in the improved deep learning model to obtain feature maps of different scales; pre-processes the feature maps of different scales through the feature pyramid network in the improved deep learning model to obtain multi-scale feature maps; performs convolution processing on the multi-scale feature maps through the head network in the improved deep learning model to generate a prediction tensor; performs confidence judgment based on the prediction tensor to obtain a fingerprint recognition result.

[0100] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0101] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0102] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0103] The readable storage medium provided by the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned fingerprint recognition method based on the improved deep learning model, and can solve the technical problem of low accuracy of the existing common fingerprint recognition technology. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present application are the same as the beneficial effects of the fingerprint recognition method based on the improved deep learning model provided by the above-mentioned embodiment, and will not be repeated here.

[0104] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the fingerprint recognition method based on the improved deep learning model as described above.

[0105] The computer program product provided by the present application can solve the technical problem of low accuracy of existing common fingerprint recognition technology. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as the beneficial effects of the fingerprint recognition method based on the improved deep learning model provided by the above embodiment, which will not be repeated here.

[0106] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A fingerprint recognition method based on an improved deep learning model, characterized in that: The improved deep learning model includes: a backbone network, a feature pyramid network and a head network, wherein the backbone network includes a deep separable convolutional layer and a partial attention module, and the fingerprint recognition method based on the improved deep learning model includes: Acquiring fingerprint information, and generating an image according to the fingerprint information; Inputting the image into the improved deep learning model, extracting features of the image through the depth-separable convolutional layer and a partial attention module in the improved deep learning model, and obtaining feature maps of different scales; Preprocessing the feature maps of different scales through a feature pyramid network in the improved deep learning model to obtain a multi-scale feature map; Performing convolution processing on the multi-scale feature map through the head network in the improved deep learning model to generate a prediction tensor; A confidence judgment is performed based on the predicted tensor to obtain a fingerprint recognition result.

2. The method according to claim 1, characterized in that The step of extracting features from the image through the deep separable convolutional layer and the partial attention module in the improved deep learning model to obtain feature maps of different scales includes: Based on the deep separable convolution layer in the improved deep learning model, a corresponding deep convolution kernel is prepared according to each channel in the image, and each channel in the image is convolved with the corresponding deep convolution kernel channel by channel, so as to obtain output features; Preparing a convolution kernel of a preset size through the depth-separable convolution layer in the improved deep learning model, performing point-by-point convolution through the convolution kernel of the preset size, and combining the output features; The combined output features are divided, connected and fused through the partial attention module, so as to output feature maps of different scales.

3. The method according to claim 1, characterized in that The step of preprocessing the feature maps of different scales through the feature pyramid network in the improved deep learning model to obtain a multi-scale feature map includes: Sampling the feature maps of different scales through a feature pyramid network in the improved deep learning model to perform feature fusion on the feature maps of different scales; The feature maps of different scales after sampling are spliced ​​so that the spliced ​​feature maps contain information of multiple scales, thereby obtaining a multi-scale feature map.

4. The method according to claim 1, characterized in that The step of performing convolution processing on the multi-scale feature map through the head network in the improved deep learning model to generate a prediction tensor comprises: Performing channel adjustment on the multi-scale feature map through the head network in the improved deep learning model so that the number of channels of the multi-scale feature map meets a preset value; A prediction tensor is generated according to the channel-adjusted multi-scale feature map, where the prediction tensor includes at least bounding box coordinates, object confidence, and category confidence.

5. The method according to claim 4, characterized in that The step of performing confidence judgment according to the predicted tensor to obtain a fingerprint recognition result comprises: Filter overlapping bounding boxes according to the bounding box coordinates, the multi-scale feature map comprising a plurality of bounding boxes; Based on the non-maximum suppression technology, confidence judgment is performed according to the predicted tensor to further filter the overlapping bounding boxes; Extract the first feature information in the filtered bounding box; Obtaining a first similarity by comparing the first feature information with preset feature information; If the first similarity is higher than a third preset threshold, it is determined that the fingerprint match is successful.

6. The method according to claim 5, characterized in that The step of performing confidence judgment based on the prediction tensor and further filtering overlapping bounding boxes based on the non-maximum suppression technology includes: Based on the non-maximum suppression technology, judging whether to filter the corresponding bounding box according to the object confidence and the category confidence; If the object confidence is lower than a first preset threshold, filtering the corresponding bounding box; If the category confidence is lower than a second preset threshold, the corresponding bounding box is filtered.

7. The method according to claim 5, characterized in that After the step of obtaining the first similarity by comparing the first feature information with the preset feature information, the method further includes: If the first similarity is not higher than a third preset threshold, it is determined that the fingerprint matching fails; Get the fingerprint module information of the terminal; Obtaining second feature information according to the fingerprint module information; Obtaining a second similarity by comparing the second feature information with the preset feature information; If the second similarity is higher than a fourth preset threshold, it is determined that the fingerprint match is successful.

8. A fingerprint recognition device based on an improved deep learning model, characterized in that: The device comprises: An acquisition module, used to acquire fingerprint information and generate an image according to the fingerprint information; A backbone network module, used to input the image into the improved deep learning model, extract features of the image through the depth-separable convolutional layer and the partial attention module in the improved deep learning model, and obtain feature maps of different scales; A feature pyramid network module, used for preprocessing the feature maps of different scales through the feature pyramid network in the improved deep learning model to obtain a multi-scale feature map; A head network module, used for performing convolution processing on the multi-scale feature map through the head network in the improved deep learning model to generate a prediction tensor; The recognition module is used to make a confidence judgment based on the predicted tensor to obtain a fingerprint recognition result.

9. A fingerprint recognition device based on an improved deep learning model, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the fingerprint recognition method based on the improved deep learning model as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by the processor, the steps of the fingerprint recognition method based on the improved deep learning model as described in any one of claims 1 to 7 are implemented.

Images