Image recognition processing methods, devices, electronic equipment and storage media

Abstract

This application provides an image recognition processing method, apparatus, electronic device, and storage medium, relating to the field of computer technology. The method includes: acquiring an image to be recognized and acquiring multiple feature matching algorithms, wherein the feature matching algorithms are adapted to the attribute information of a target object in the image to be recognized; extracting image features from the image to be recognized according to the multiple feature matching algorithms to obtain multiple image features; fusing the multiple image features to obtain target features; and recognizing the target object in the image to be recognized based on the target features to obtain a recognition result. This application addresses the problem of low recognition success rates in related technologies such as biometric recognition.

Description

Technical Field

[0001] This application relates to the field of computer technology, and more specifically, to an image recognition processing method, apparatus, electronic device, and storage medium. Background Technology

[0002] With the development of computer technology, image recognition processing has been widely applied in various fields, such as image-based biometric recognition. For example, in the field of public safety, biometric recognition can be facial recognition. For a registered user of a building, if facial recognition of the registered user is successful, the registered user can successfully pass through the building's access control.

[0003] Currently, the success rate of biometric identification depends not only on the image being identified—for example, higher image quality improves the success rate—but also on the biometric features themselves. Taking fingerprint recognition as an example, the quality of fingerprints varies significantly among different populations; those with poor fingerprint quality have a lower success rate compared to those with good fingerprint quality. Furthermore, electronic devices used for biometric identification are susceptible to external factors, such as access control devices that are easily affected by seasonal changes and human wear and tear, thus impacting the success rate of biometric identification.

[0004] Therefore, improving the success rate of biometric identification remains a challenge. Summary of the Invention

[0005] This application provides an image recognition processing method, apparatus, electronic device, and storage medium, which can solve the problem of low recognition success rate in related technologies, such as biometric recognition. The technical solution is as follows: According to one aspect of the embodiments of this application, an image recognition processing method is provided, the method comprising: acquiring an image to be recognized and acquiring multiple feature matching algorithms, wherein the feature matching algorithms are adapted to attribute information of a target object in the image to be recognized; extracting image features from the image to be recognized according to the multiple feature matching algorithms to obtain multiple image features; fusing the multiple image features to obtain target features; and recognizing the target object in the image to be recognized according to the target features to obtain a recognition result.

[0006] According to one aspect of the embodiments of this application, an image recognition processing apparatus includes: an algorithm acquisition module for acquiring an image to be recognized and acquiring multiple feature matching algorithms, wherein the feature matching algorithms are adapted to attribute information of a target object in the image to be recognized; a feature extraction module for extracting image features from the image to be recognized according to the multiple feature matching algorithms to obtain multiple image features; a feature fusion module for fusing the multiple image features to obtain target features; and a feature recognition module for recognizing the target object in the image to be recognized according to the target features to obtain a recognition result.

[0007] In an exemplary embodiment, the algorithm acquisition module includes: an information extraction unit, configured to extract the attribute information from the image to be identified; and an algorithm selection unit, configured to select from an algorithm set a variety of feature matching algorithms that are compatible with the attribute information, wherein the algorithm set includes a variety of candidate algorithms for image feature extraction that are available for selection.

[0008] In an exemplary embodiment, the algorithm selection unit includes: an algorithm search subunit, configured to search for various candidate algorithms in the algorithm set based on the attribute information; and an algorithm adaptation subunit, configured to select candidate algorithms whose adaptation degree meets the adaptation condition as the feature matching algorithm based on the degree of adaptation between the candidate algorithms and the attribute information.

[0009] In an exemplary embodiment, the attribute information includes a primary attribute and a secondary attribute associated with the primary attribute; the apparatus further includes: a set construction module for constructing the algorithm set; the set construction module includes: an algorithm classification unit for classifying multiple candidate algorithms according to different primary attributes to obtain multiple algorithm categories, wherein the candidate algorithms in each algorithm category correspond to the same primary attribute; an algorithm adaptation unit for adapting the candidate algorithms in each algorithm category to the secondary attribute associated with their corresponding primary attribute; and a path construction unit for constructing a path between the adapted candidate algorithms and the secondary attribute, and configuring the adaptation degree for the constructed path to obtain the algorithm set.

[0010] In an exemplary embodiment, the algorithm search subunit includes a correspondence subunit, configured to search for candidate algorithms that correspond to the secondary attributes in the algorithm set based on the correspondence between the secondary attributes and candidate algorithms in the algorithm set.

[0011] In an exemplary embodiment, the feature recognition module includes: a feature segmentation unit, configured to segment the target features to obtain multiple target feature sub-segments; a feature recognition unit, configured to perform biometric recognition on the target object in the image to be recognized according to each target feature group, to obtain a recognition result corresponding to each target feature group, wherein the target feature group includes a set number of target feature sub-segments from the multiple target feature sub-segments; and a result generation unit, configured to obtain the recognition result according to the recognition result corresponding to each target feature group.

[0012] In an exemplary embodiment, the feature recognition unit includes: a sample acquisition subunit, configured to acquire a sample feature group corresponding to each target feature group for each sample feature in the sample set, wherein the sample feature group includes a predetermined number of sample feature segments from a plurality of sample feature segments, and the sample feature segments are obtained by segmentation processing of the sample features; and a similarity calculation subunit, configured to calculate the similarity between each target feature group and the acquired sample feature group, and use it as the recognition result corresponding to each target feature group.

[0013] In one exemplary embodiment, the target object includes a fingerprint.

[0014] According to one aspect of the embodiments of this application, an electronic device includes: at least one processor, at least one memory, and at least one communication bus, wherein a computer program is stored in the memory, and the processor reads the computer program in the memory through the communication bus; when the computer program is executed by the processor, it implements the image recognition processing method as described above.

[0015] According to one aspect of the embodiments of this application, a storage medium stores a computer program thereon, which, when executed by a processor, implements the image recognition processing method as described above.

[0016] According to one aspect of the embodiments of this application, a computer program product includes a computer program stored in a storage medium. A processor of a computer device reads the computer program from the storage medium and executes the computer program, causing the computer device to implement the image recognition processing method as described above when executed.

[0017] The beneficial effects of the technical solution provided in this application are: In the above technical solution, an image to be identified and multiple feature matching algorithms adapted to the attribute information of the target object in the image to be identified are obtained. Image features are extracted from the image to be identified according to these feature matching algorithms to obtain multiple image features. The target features are obtained by feature fusion of multiple image features. Finally, the target object in the image to be identified is identified according to the target features to obtain the identification result. Taking fingerprint recognition as an example, for people with poor fingerprint quality and people with good fingerprint quality, the quality of fingerprints in the images to be identified is different. Therefore, the multiple feature matching algorithms adapted to different fingerprint qualities of the images to be identified are also different. In other words, through adaptation, more effective feature matching algorithms will be obtained for images to be identified with different fingerprint qualities. This makes the biometric identification success rate of the images to be identified using the feature matching algorithm higher, thereby effectively solving the problem of low recognition success rate in related technologies such as biometric identification. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below.

[0019] Figure 1 This is a schematic diagram of the implementation environment according to the embodiments of this application; Figure 2 This is a flowchart illustrating an image recognition processing method according to an exemplary embodiment; Figure 3 yes Figure 2 A flowchart of step 350 in one embodiment corresponds to the following example; Figure 4 yes Figure 2 A flowchart of step 310 in one embodiment corresponds to the following example; Figure 5 yes Figure 4 A flowchart of step 313 in one embodiment corresponds to the following example; Figure 6 This is a flowchart illustrating another image recognition processing method according to an exemplary embodiment; Figure 7 yes Figure 6 A schematic diagram of the set of algorithms involved in the corresponding embodiments in one embodiment; Figure 8 yes Figure 2 A flowchart of step 370 in one embodiment corresponds to the following example; Figure 9 This is a schematic diagram illustrating the specific implementation of an image recognition processing method in an application scenario; Figure 10 yes Figure 9A schematic diagram of the fingerprint feature extraction architecture involved in the application scenario; Figure 11 This is a structural block diagram of an image recognition processing apparatus according to an exemplary embodiment; Figure 12 This is a hardware structure diagram of an electronic device according to an exemplary embodiment; Figure 13 This is a structural block diagram of an electronic device according to an exemplary embodiment. Detailed Implementation

[0020] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0021] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this application means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0022] The following is an introduction and explanation of several terms used in this application: FRR, short for False Reject Rate, refers to the probability that a registered user's biometric features are not successfully identified. A registered user is one whose biometric features are stored in a biometric database. Conversely, the pass rate (1 - False Reject Rate) is the opposite of the pass rate, representing the probability that a registered user's biometric features are successfully identified.

[0023] FAR, short for False Accepted Rate, is the probability that a person's biometric features are mistakenly identified as those of a registered user in a biometric database, thus allowing the user to pass through. "Others" refers to users whose biometric features are not stored in the database.

[0024] As mentioned earlier, the success rate of biometric recognition is affected not only by the image to be recognized, but also by the biometric features of the target itself. In addition, the electronic devices used for biometric recognition can also have a low success rate due to external factors.

[0025] Currently, relevant technologies have proposed improvement schemes for images to be identified. Specifically, this involves using one or more methods such as image preprocessing, image quality assessment, and image feature adjustment to control the quality of the images to be identified, thereby improving the success rate of biometric identification.

[0026] However, although the above-mentioned improvement schemes can improve the image quality of the image to be identified to a certain extent, the success rate of biometric identification caused by the biometric features of the target itself and electronic devices is low, and the improvement effect is not obvious. For example, when there are large differences in fingerprint quality among different groups of people and when different external factors cause large changes in fingerprints, the improvement effect on the image quality of the image to be identified is not good.

[0027] As can be seen from the above, the relevant technologies still suffer from the drawback of low success rate in biometric identification.

[0028] Therefore, the image recognition processing method provided in this application can effectively improve the success rate of target recognition. Accordingly, the image recognition processing method is applicable to image recognition processing devices, which can be deployed in electronic devices, such as image acquisition devices, desktop computers, laptops, tablets, servers, etc., which have biometric recognition functions.

[0029] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0030] Figure 1 This is a schematic diagram of the implementation environment involved in an image recognition processing method. As shown in Figure (a), the implementation environment includes a gateway 110, an image acquisition device 130 deployed in the gateway 110, and a server 150.

[0031] The image acquisition device 130 can be a camera, a smartphone, a tablet computer, or other electronic devices equipped with a camera. It can also be an electronic device with the function of acquiring specific image textures, such as a smart door lock. There is no limitation here.

[0032] Server 150 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. For example, in this implementation environment, biometric identification services are provided by server 150.

[0033] Image acquisition device 130 is deployed on gateway 110 and communicates with gateway 110 through its own configured communication module, thereby realizing interaction with gateway 110. In one application scenario, image acquisition device 130 is deployed on gateway 110 by accessing it through a local area network (LAN). The process of image acquisition device 130 accessing gateway 110 through the LAN includes gateway 110 first establishing a LAN, and image acquisition device 130 connecting to the LAN established by gateway 110 to access the LAN established by gateway 110. This LAN includes, but is not limited to, Bluetooth, WIFI, ZigBee, or LoRa.

[0034] Server 150 establishes a communication connection with gateway 110 beforehand, and uses this communication connection to realize data transmission between server 150 and gateway 110. For example, the transmitted data includes at least the image to be identified.

[0035] In one application scenario, the image acquisition device 130 captures and collects an image to be identified, and transmits the image to be identified to the server 150 through the gateway 110, so that the server 150 can provide biometric identification services.

[0036] For server 150, after receiving the image to be identified, it can call the biometric identification service for the image to be identified. Specifically, it can: acquire multiple feature matching algorithms that are adapted to the attribute information of the target object in the image to be identified; extract image features from the image to be identified according to these feature matching algorithms to obtain multiple image features; and obtain the target features by feature fusion of multiple image features. Finally, it can identify the target object in the image to be identified according to the target features to obtain the identification result, thereby solving the problem of low recognition success rate of biometric identification in related technologies.

[0037] Of course, in other application scenarios, for image acquisition devices that also have biometric recognition capabilities, after capturing and acquiring the image to be recognized, they can directly identify the target object in the image, thereby obtaining the recognition result for relevant applications. For example, the image acquisition device can be an access control device with fingerprint recognition capabilities, allowing visitors to pass through the access control smoothly by using fingerprint recognition.

[0038] Unlike Figure (a), in Figure (b), the implementation environment also includes user terminal 170.

[0039] Specifically, the user terminal 170, which can also be considered a user client or terminal, can be run by a client with target display capabilities. This user terminal 170 can be an electronic device such as a desktop computer, laptop computer, tablet computer, or smartphone, and is not limited here. The client, having target display capabilities, can be in the form of an application or a webpage. Correspondingly, the interface for target display on the client can be in the form of a program window or a webpage, and is not limited here either.

[0040] A communication connection is pre-established between the user terminal 170 and the server 150, and data transmission between the user terminal 170 and the server 150 is realized through this communication connection. For example, the transmitted data may be identification results, etc.

[0041] In one application scenario, user terminal 170 initiates a biometric recognition request to server 170 via a running client, requesting server 170 to provide biometric recognition services for the image to be recognized. After server 170 obtains the recognition result by calling the biometric recognition service, it can return the recognition result to user terminal 170 for display on user terminal 170 based on the recognition result.

[0042] Of course, in other application scenarios, user terminals equipped with cameras can also integrate functions such as biometric recognition and display. Taking face recognition as an example, after the user terminal captures and collects the image to be recognized, it can directly recognize the face in the image and display the recognized face on the corresponding interface, or perform related applications based on the recognized face, such as unlocking smartphones with face recognition.

[0043] Please see Figure 2 This application provides an image recognition processing method, which is applicable to electronic devices, specifically, [the electronic devices may be...]. Figure 1 The image acquisition device 130, server 150, user terminal 170, etc. are shown in the implementation environment.

[0044] In the following method embodiments, for ease of description, the execution subject of each step of the method is an electronic device, but this does not constitute a specific limitation.

[0045] like Figure 2 As shown, the method may include the following steps: Step 310: Obtain the image to be recognized and acquire multiple feature matching algorithms that are adapted to the attribute information of the target object in the image to be recognized.

[0046] First, it should be clarified that the image to be identified is obtained by capturing and collecting images of a target. The target corresponds to the object in the image to be identified. In other words, biometric identification is performed on the target object in the image to be identified. Therefore, the target can refer to physiological characteristics such as a person's face, fingerprints, palm prints, iris, earlobes, and pulse, or behavioral characteristics such as gait, handwriting, and voice, so that image-based biometric identification methods can be applied to different application scenarios. For example, in a public safety application scenario, visitors can pass through access control smoothly using fingerprint recognition; in another criminal investigation application scenario, facial recognition can be used to track an individual.

[0047] It is understood that shooting can be a single shot or continuous shooting. For the same target, continuous shooting yields a video clip, and the image to be identified can originate from one frame of that video. Conversely, a single shot yields multiple photos, and the image to be identified can originate from one of those photos. In other words, biometric identification in this embodiment is performed frame-by-frame.

[0048] Regarding the acquisition of the image to be recognized, the image can originate from images acquired in real time by an image acquisition device, or it can be images acquired by the image acquisition device within a historical time period that are pre-stored on a server. In other words, after the image acquisition device acquires the image to be recognized, the electronic device can process the image in real time, or it can pre-store it for later processing, for example, processing it according to a time specified by the operator. Therefore, the electronic device can acquire images acquired in real time, or it can acquire pre-stored images, i.e., by retrieving images acquired within a historical time period; this embodiment does not limit this choice.

[0049] In one possible implementation, the target object includes biometric features, which include physiological and behavioral characteristics. For example, the target object can be a person's fingerprint, a person's face, etc. Accordingly, the attribute information can also be different depending on the target object. For example, if the target object is a face, the attribute information can be gender, face shape, etc., or if the target object is a fingerprint, the attribute information can be fingerprint quality, fingerprint type, etc. This embodiment does not limit the attribute information of the target object.

[0050] Secondly, "adaptation" refers to the high success rate of biometric identification when using feature matching algorithms on the image to be identified. In other words, an adapted feature matching algorithm can more effectively improve the success rate of subsequent biometric identification.

[0051] Regarding the acquisition of suitable feature matching algorithms, one possible implementation involves obtaining multiple feature matching algorithms adapted to the attribute information of the target object in the image to be identified. The attribute information can indicate the identifiability of the target object in the image to be identified, or it can indicate the type of the target object. This attribute information can be extracted by the electronic device from the image to be identified, or it can be generated by the image acquisition device through configuration when the image to be identified is captured and sent to the electronic device; no limitation is imposed here.

[0052] It should be noted that different levels of identifiability essentially refer to different probabilities of successfully identifying the target object in the image to be identified. The higher the identifiability, the higher the probability of successfully identifying the target object in the image to be identified. Therefore, as the identifiability varies, the feature matching algorithm adapted to the attribute information will also differ. For example, for an image to be identified that contains a target object with low identifiability, a feature matching algorithm that can more effectively improve the success rate of biometric identification will be adapted. This will enable more effective biometric identification for different images to be identified, such as those with large differences in image quality and large changes in the target due to different external factors, thereby improving the success rate of biometric identification.

[0053] Taking the target object in the image to be identified as the fingerprint and the attribute information as the fingerprint quality as an example, the feature matching algorithms include, but are not limited to: algorithms for extracting minutiae features, algorithms for extracting fine line structure features, algorithms for extracting global features, algorithms for extracting local features, algorithms for extracting topological structure features, algorithms for extracting mixed features of fine line structure features and minutiae features, triangulation algorithms, triangulation matching algorithms, orientation field repair algorithms, etc.

[0054] Therefore, for fingerprint images with good quality to be identified, suitable feature matching algorithms include, but are not limited to: algorithms for extracting minutiae features, triangulation algorithms, and triangulation matching algorithms.

[0055] For fingerprint images with poor quality, suitable feature matching algorithms include, but are not limited to: algorithms for extracting local features, algorithms for extracting topological features, etc.

[0056] It is worth mentioning that the fingerprint quality attribute can be represented by a configured quality score. For example, a good fingerprint quality is represented by a quality score of 70 to 100, which indicates that the fingerprint in the image to be identified has a high degree of recognizability; a poor fingerprint quality is represented by a quality score of less than 70, which indicates that the fingerprint in the image to be identified has a low degree of recognizability. This is not a specific limitation.

[0057] Step 330: Based on multiple feature matching algorithms, image features are extracted from the image to be identified to obtain multiple image features.

[0058] Each image feature corresponds to a feature matching algorithm. This image feature is an accurate description of the target object in the image to be identified. It should be understood that different target objects in the image will result in different extracted image features; in other words, the image feature uniquely identifies the target object in the image.

[0059] After determining various feature matching algorithms, multiple corresponding image features can be identified through image feature extraction. These multiple image features can be understood as accurately describing the target object in the image to be identified from different dimensions.

[0060] Taking fingerprints as an example, minutiae features can be extracted from the image to be identified using an algorithm for extracting minutiae features. These minutiae features accurately describe the minutiae of the fingerprint in the image to be identified. Similarly, fine line structure features can be extracted from the image to be identified using an algorithm for extracting fine line structure features. These fine line structure features accurately describe the texture of the fingerprint in the image to be identified.

[0061] Step 350: Perform feature fusion on multiple image features to obtain the target features.

[0062] Feature fusion includes, but is not limited to, consistency processing, binarization processing, and normalization processing. It should be noted that the inventors recognized that since multiple image features are obtained through various feature matching algorithms, their feature vector dimensions are not entirely the same. Therefore, consistency processing refers to unifying the different feature vector dimensions of multiple image features.

[0063] In one possible implementation, such as Figure 3 As shown, step 350 may include the following steps: step 351, performing consistency processing on multiple image features with respect to feature vector dimensions to obtain intermediate features; step 353, performing binarization processing on the intermediate features to obtain target features.

[0064] Feature fusion can be achieved using fully connected networks or other machine learning methods; no specific limitations are specified here.

[0065] In this approach, by fusing target features obtained from multiple image features, the target object in the image to be identified can be accurately described from different dimensions simultaneously, which helps to improve the robustness of biometric identification and thus improve the success rate of biometric identification.

[0066] Step 370: Identify the target object in the image to be identified based on the target features to obtain the identification result.

[0067] The target object can be identified through methods such as target classification and target retrieval, without any specific limitations. Correspondingly, the identification result is used to indicate the category to which the target object belongs, or to indicate a sample that matches the target object, or to indicate whether biometric identification was successful, etc.

[0068] In one possible implementation, the recognition process based on target classification may include the following steps: predicting the category of the target object in the image to be recognized based on the target features to obtain the predicted category of the target object; and generating a recognition result based on the predicted category of the target object.

[0069] For example, the target object is a human face as the biometric feature to be identified. Faces are used in the target classification process. Assume that face categories include female faces and male faces. Category prediction essentially involves calculating the probability P1 of the face to be identified being a female face and the probability P2 of it being a male face, based on the facial features.

[0070] If the probability P1 of a face being female is greater than or equal to the probability P2 of a face being male, then the face to be identified is predicted to be a female face, that is, the predicted category of the face to be identified is female face. Conversely, if the probability P of a face being female is less than the probability P2 of a face being male, then the face to be identified is predicted to be a male face, that is, the predicted category of the face to be identified is male face.

[0071] Furthermore, the predicted category of the face to be identified can be used directly as the identification result, or it can be used as the identification result after meeting the set conditions for biometric identification. It can be understood that the set conditions here mean that the probability of the face to be identified belonging to the face category is greater than the probability threshold. For example, if the probability threshold is 0.9, then the face category to which the face to be identified belongs can be used as the identification result if P1 / P2 is greater than 0.9.

[0072] In one possible implementation, the recognition process based on target retrieval may include the following steps: traversing the sample features of each sample image in the sample library, calculating the similarity between the target features and the sample features of each sample image; and generating a recognition result based on the calculated similarity.

[0073] For example, the same principle applies to the recognition process of target retrieval. Suppose that the sample database stores sample features A1, B1, and C1 of registered users A, B, and C. Then, we can iterate through all the sample features in the sample database and calculate the similarity between the face features and the sample features A1, B1, and C1 respectively.

[0074] Furthermore, the face of the registered user corresponding to the sample feature with the highest similarity can be selected as the recognition result. Alternatively, after the maximum similarity meets the set conditions for biometric recognition, the face of the corresponding registered user can be used as the recognition result. Or, after all similarities fail to meet the set conditions for biometric recognition, a recognition result indicating face recognition failure can be generated. It can be understood that the set conditions here refer to similarity greater than a similarity threshold. For example, if the similarity threshold is 0.9, and the similarity between the face feature and sample features A1, B1, and C1 is all less than 0.9, then the recognition result is used to indicate face recognition failure.

[0075] Through the above process, by adapting the feature matching algorithm to the image to be identified, more effective feature matching algorithms will be obtained for different images to be identified, so that the biometric identification of the image to be identified using the feature matching algorithm has a higher success rate, thereby effectively solving the problem of low success rate of biometric identification in related technologies.

[0076] Please see Figure 4 In one exemplary embodiment, step 310 may include the following steps: Step 311: Extract attribute information from the image to be identified.

[0077] Among them, attribute information is used to indicate the recognizability of target objects in the image to be recognized, and can also be used to indicate the type of target objects in the image to be recognized, etc.

[0078] Taking fingerprints as an example, the attribute information includes, but is not limited to: fingerprint type, fingerprint quality, number of minutiae, minutiae type, number of pseudo-minutiae, fingerprint dryness / moistness, image quality, image distortion, etc. It can be understood that, on the one hand, good fingerprint quality, a large number of minutiae, a large number of pseudo-minutiae, a dry fingerprint, good image quality, and no image distortion all indicate high fingerprint recognizability in the image to be identified; conversely, poor fingerprint quality, a small or missing number of minutiae, a small or missing number of pseudo-minutiae, a moist fingerprint, poor image quality, and image distortion all indicate low fingerprint recognizability in the image to be identified. On the other hand, fingerprint type and minutiae type indicate the type of fingerprint in the image to be identified.

[0079] Step 313: Select from the algorithm set a variety of feature matching algorithms that are compatible with the attribute information.

[0080] The algorithm set includes at least a selection of candidate algorithms for image feature extraction.

[0081] For fingerprints as the target object, the algorithm set can be constructed from at least two of the following candidate algorithms: algorithms for extracting minutiae features, algorithms for extracting fine line structure features, algorithms for extracting global features, algorithms for extracting local features, algorithms for extracting topological structure features, algorithms for extracting hybrid features of fine line structure features and minutiae features, triangulation algorithms, triangulation matching algorithms, orientation field repair algorithms, etc.

[0082] Regarding the selection of the appropriate feature matching algorithm, one possible implementation is to randomly select an appropriate feature matching algorithm from the algorithm set; another possible implementation is to select a candidate algorithm from the algorithm set whose degree of fit with the attribute information meets the fitting condition.

[0083] It should be noted that the degree of fit refers to the extent to which the feature matching algorithm matches the attribute information in the image to be identified. In other words, the higher the degree of fit, the better the feature matching algorithm matches the attribute information in the image, and the more effectively it can improve the success rate of biometric recognition. This degree of fit can be represented by the weight values ​​configured for each candidate algorithm during the construction of the algorithm set, or by the false rejection rate and / or rejection rate obtained by testing each candidate algorithm on the same test set. The purpose of the fit criteria is to select feature matching algorithms that can more effectively improve the success rate of biometric recognition. This fit criteria can refer to the top K feature matching algorithms in terms of fit (e.g., K=2), or it can refer to feature matching algorithms whose fit exceeds the fit threshold. Both the degree of fit and the fit criteria can be flexibly set according to the actual needs of the application scenario, and are not limited here.

[0084] Taking fingerprints as an example, for an image to be identified with attributes such as good fingerprint quality, a large number of minutiae, and dry fingerprints, the appropriate feature matching algorithm can be at least two of the following: an algorithm for extracting minutiae features, a triangulation algorithm, a triangulation matching algorithm, etc.

[0085] For images to be identified with attributes such as poor fingerprint quality, few or missing minutiae, many false minutiae, or wet fingerprints, the appropriate feature matching algorithm can be at least two of the following: an algorithm for extracting local features, an algorithm for extracting topological features, etc.

[0086] For images to be identified with poor image quality and large image distortion, the appropriate feature matching algorithm can be at least two of the following: algorithms for extracting local features, orientation field repair algorithms, etc.

[0087] With the cooperation of the above embodiments, the attribute information extracted from the image to be identified provides a basis and support for the adaptation of the feature matching algorithm, which is conducive to improving the robustness of biometric identification and thus improving the success rate of biometric identification.

[0088] Please see Figure 5 In one exemplary embodiment, step 313 may include the following steps: Step 3131: Based on the attribute information, search for multiple candidate algorithms in the algorithm set.

[0089] In other words, the algorithm set not only includes multiple candidate algorithms for image feature extraction, but also the correspondence between attribute information and candidate algorithms. Therefore, based on the correspondence between attribute information and candidate algorithms in the algorithm set, it is possible to find candidate algorithms that correspond to the attribute information in the algorithm set and use them as feature matching algorithms.

[0090] In one possible implementation, the attribute information includes primary attributes and secondary attributes associated with the primary attributes. The primary attribute refers to a criterion that defines the type of target object in the image to be identified, such as fingerprint type; the secondary attribute is a sub-criterion belonging to the criterion, such as the arch pattern in fingerprint type. Alternatively, the primary attribute can be a criterion used to measure the identifiability of the target object in the image to be identified, such as fingerprint quality; the secondary attribute is a sub-criterion belonging to the criterion, such as good fingerprint quality or poor fingerprint quality in fingerprint quality.

[0091] As mentioned earlier, for the target object being a fingerprint, the attribute information includes, but is not limited to: fingerprint type, fingerprint quality, number of minutiae, minutiae type, number of pseudo-minutiae, fingerprint dryness / moistness, image quality, image distortion, etc.

[0092] Based on this, for fingerprints as the target object, the primary attributes include, but are not limited to: fingerprint type, fingerprint quality, number of minutiae, minutiae type, number of pseudo-minutiae, fingerprint dryness / wetness, image quality, image distortion, etc.

[0093] Therefore, the related secondary attributes for the aforementioned primary attributes specifically refer to the following: Secondary attributes related to fingerprint type include: arch, loop, and whorl patterns; secondary attributes related to fingerprint quality include: good fingerprint quality and poor fingerprint quality; secondary attributes related to the number of minutiae include: many minutiae and few minutiae; secondary attributes related to minutiae type include: center point, triangular point, termination point (also considered as endpoint), bifurcation point (also considered as intersection), divergence point, and isolated point; secondary attributes related to the number of pseudo-minutiae include: many pseudo-minutiae and few pseudo-minutiae; secondary attributes related to fingerprint dryness / moistness include: dry fingerprint and moist fingerprint; secondary attributes related to image quality include: good image quality and poor image quality; and secondary attributes related to image distortion include: large image distortion and small image distortion.

[0094] Therefore, in one possible implementation, multiple candidate algorithms are searched in the algorithm set based on the primary attribute, and the algorithm set includes at least the correspondence between the primary attribute and the candidate algorithm; in another possible implementation, multiple candidate algorithms are searched in the algorithm set based on the secondary attribute, and the algorithm set includes at least the correspondence between the secondary attribute and the candidate algorithm.

[0095] like Figure 6 As shown, in one possible implementation, the construction of the algorithm set may include the following steps: Step 410, classifying multiple candidate algorithms according to different primary attributes to obtain multiple algorithm categories, with each candidate algorithm in the algorithm category corresponding to the same primary attribute; Step 430, for each candidate algorithm in the algorithm category, adapting the candidate algorithm to the secondary attribute associated with its corresponding primary attribute; Step 450, constructing a path between the adapted candidate algorithm and the secondary attribute, and configuring the degree of adaptation for the constructed path to obtain the algorithm set.

[0096] Figure 7 A schematic diagram of an algorithm set is shown. Figure 7In this model, the primary attributes include fingerprint type, minutiae type, number of minutiae (61), fingerprint dryness / moistness, and image distortion (63). The secondary attributes associated with these primary attributes vary. For example, the secondary attribute (611) associated with the number of minutiae (611) includes whether the number of minutiae is high or low; the secondary attribute (631) associated with image quality (633) includes whether the image distortion is high or low. Based on these primary attributes and their associated secondary attributes, multiple candidate algorithms can be classified according to different primary attributes, resulting in multiple algorithm categories. For example, candidate algorithms in algorithm category 612 correspond to the number of minutiae (61), and candidate algorithms in algorithm category 632 correspond to image distortion (63). Even within the same algorithm category, candidate algorithms address different problems. For example, algorithms for extracting local features can solve problems with large image distortion, while algorithms for extracting global features are suitable for cases with small image distortion. In other words, different candidate algorithms may be adapted to different secondary attributes. Therefore, for each algorithm category, it is necessary to adapt the candidate algorithm to the secondary attributes. For instance, for multiple candidate algorithms in algorithm category 632, those adapted to large image distortion include algorithms for extracting local features and orientation field restoration algorithms, while those adapted to small image distortion include algorithms for extracting detailed power loss features and algorithms for extracting global features.

[0097] After matching candidate algorithms with secondary attributes, a path can be constructed between the matched candidate algorithms and secondary attributes. This can also be understood as the path indicating the matched candidate algorithms and secondary attributes, or indicating the candidate algorithms and secondary attributes that have a corresponding relationship. The degree of adaptation is configured for the constructed path. For example, path 613 and its degree of adaptation 1, path 633 and its degree of adaptation 2, thus obtaining the algorithm set 600.

[0098] Taking the secondary attribute in the attribute information as an example, with a detail quantity of 611, such as... Figure 7 As shown, based on the correspondence between secondary attributes and candidate algorithms in the algorithm set 600, and considering the large number of details in the attribute information, candidate algorithms that correspond to a large number of details can be found in the algorithm set 600, including: algorithms for extracting minutiae features, triangulation algorithms, and triangulation matching algorithms. It should be noted that candidate algorithms that correspond to secondary attributes can also be considered as candidate algorithms that have a path established with the secondary attributes, that is, candidate algorithms that are adapted to the secondary attributes.

[0099] Step 3133: Based on the degree of fit between the candidate algorithms and the attribute information, the candidate algorithms that meet the fit conditions are selected as feature matching algorithms.

[0100] Continue reading Figure 7As can be seen, each candidate algorithm that corresponds to the secondary attribute has a corresponding degree of fit. Therefore, after determining the candidate algorithms that correspond to the secondary attribute, the candidate algorithms found in step 3131 can be filtered according to their degree of fit, so that the candidate algorithms whose degree of fit meets the fit condition are used as feature matching algorithms.

[0101] In one possible implementation, the top three candidate algorithms in terms of fit are considered to meet the fit criteria.

[0102] In one possible implementation, the fit is the false rejection rate and / or rejection rate obtained by testing each candidate algorithm on the same test set.

[0103] Under the above embodiments, by classifying and evaluating various candidate algorithms, an algorithm set is constructed, providing a basis and support for adapting more effective feature matching algorithms to different images to be identified, thereby helping to improve the success rate of biometric identification.

[0104] Please see Figure 8 In one exemplary embodiment, step 370 may include the following steps: Step 371: The target features are segmented to obtain multiple target feature sub-segments.

[0105] For example, the feature vector used to represent the target feature can be divided into n equal segments, and each segment of the feature vector can be regarded as a target feature sub-segment. Alternatively, the target feature can be considered to include n target feature sub-segments.

[0106] Step 373: Traverse multiple target feature groups, perform biometric recognition on the target object in the image to be recognized based on the currently traversed target feature group, and obtain the recognition result corresponding to the currently traversed target feature group.

[0107] As mentioned earlier, target objects can be identified through target classification, target retrieval, and other methods. The inventors realized that, especially with the target retrieval method, since it is necessary to calculate the similarity between each sample feature in the sample set and the target feature, as the number of sample features in the sample set increases, the amount of similarity calculation will increase accordingly, which will also affect the similarity calculation speed and thus affect the recognition efficiency of biometrics.

[0108] Based on this, in this embodiment, biometric identification is performed on targets in the image to be identified by segmenting the target feature groups, thereby improving the identification efficiency of biometric identification. The target feature group includes a predetermined number of target feature sub-segments from multiple target feature sub-segments. Taking the previous example, assuming the predetermined number is r, and m = n / r, then the target features include m target feature groups, and each target feature group includes r target feature sub-segments. It is worth noting that considering n / r is not necessarily divisible, the number of target feature sub-segments contained in the last target feature group is between 1 and r.

[0109] After identifying multiple target feature groups contained within the target features, biometric identification of the target object in the image to be identified can be performed based on each target feature group. An example of the currently traversed target feature group is as follows: In one possible implementation, based on the target feature group currently traversed, the category of the target object in the image to be recognized is predicted to obtain the probability that the target object belongs to different categories; the probability that the target object belongs to different categories is used as the recognition result corresponding to the target feature group currently traversed.

[0110] In one possible implementation, for each sample feature in the sample set, a sample feature group corresponding to the currently traversed target feature group is obtained. The sample feature group includes a set number of sample feature sub-segments from multiple sample feature sub-segments, which are obtained by segmenting the sample features. The similarity between the currently traversed target feature group and the obtained sample feature group is calculated and used as the recognition result corresponding to the currently traversed target feature group.

[0111] Step 375: Determine the recognition result of the target object based on the recognition results corresponding to all target feature groups that have been traversed.

[0112] Step 377: If the identification result of the target object meets the set conditions for biometric identification, then generate the identification result based on the identification result of the target object.

[0113] Taking a target feature consisting of m target feature groups, and each target feature group consisting of r target feature sub-segments as an example, the recognition process based on target classification is explained as follows: Assuming the categories include category a and category b, for the first target feature group, if the probability that the target object belongs to category a is P1 and the probability that the target object belongs to category b is P2, this is taken as the recognition result corresponding to the first target feature group. At this time, since there are no other target feature groups that have been traversed, it can be determined that the probability Pa of the target object belonging to category a is P1 and the probability Pb ​​of the target object belonging to category b is P2.

[0114] If Pa is greater than Pb, then the target object belongs to category a, and this is taken as the identification result of the target object.

[0115] Conversely, if Pb is greater than Pa, then the target object belongs to category b, which is used as the identification result of the target object.

[0116] Assuming the biometric recognition condition is that the probability exceeds a probability threshold (e.g., the probability threshold is 0.9), then for the first target feature group, the biometric recognition condition is that the probability exceeds 0.9×r / n (the probability threshold for the r target feature segments that have been traversed). If Pa or Pb is less than 0.9×r / n, that is, the recognition result of the target object does not meet the biometric recognition condition, then the biometric recognition for the first target feature group is determined to be a failure. At this time, biometric recognition continues for the first two target feature groups.

[0117] For the second target feature group, if the probability of the target object belonging to category a is P3 and the probability of the target object belonging to category b is P4, this is taken as the recognition result corresponding to the second target feature group. Then, combining this with the recognition result corresponding to the first target feature group, for the first target feature group, the probabilities P1 and P2 of the target object belonging to category a and b respectively, can be used to determine that the probability Pa of the target object belonging to category a is α×P1+β×P3, and the probability Pb ​​of the target object belonging to category b is α×P2+β×P4. Here, α and β are the weights assigned to the first and second target feature groups, respectively.

[0118] Similarly, if Pa is greater than Pb, then the target object belongs to category a, which is used as the identification result of the target object.

[0119] Conversely, if Pb is greater than Pa, then the target object belongs to category b, which is used as the identification result of the target object.

[0120] At this point, if Pa or Pb is greater than 0.9×2r / n (the probability threshold for the 2r target feature segments that have been traversed), that is, the recognition result of the target object meets the set conditions for biometric recognition, then the biometric recognition of the first two target feature groups is determined to be successful, and the recognition result is obtained from the recognition result of the target object, that is, the target object belongs to category a.

[0121] Otherwise, continue biometric identification for the first three target feature groups until biometric identification is successful or all m target feature groups have been traversed.

[0122] In the above process, segmented feature-based biometric recognition is realized, that is, biometric recognition is performed on the image to be recognized by segmenting the target feature group. It is applicable not only to target classification-based biometric recognition, but also to target retrieval-based biometric recognition, which can effectively improve the recognition speed of biometric recognition, and thus effectively improve the recognition efficiency of biometric recognition.

[0123] Figures 9 to 10 This is a schematic diagram illustrating the specific implementation of an image recognition processing method in an application scenario. This application scenario is suitable for... Figure 1 In the implementation environment shown, when the target object is a fingerprint, the image acquisition device 130 can be an electronic device with fingerprint acquisition function, such as a fingerprint lock or fingerprint access control system, for example, a smart door lock. The image acquisition device 130 captures and acquires the fingerprint image, and forwards it to the server 150 through the gateway 110 to perform fingerprint recognition on the fingerprint image.

[0124] Now combined Figures 9 to 10 The following explains the two branches involved in the fingerprint recognition process: First, regardless of whether it is a registration branch or a retrieval branch, fingerprint feature extraction needs to be performed on the obtained fingerprint image, that is, steps 81 to 82 are executed.

[0125] Figure 10 A schematic diagram of the fingerprint feature extraction architecture is shown. Figure 10 In this fingerprint feature extraction architecture, attribute extraction 821, algorithm selection 822, and feature fusion 823 are included. Attribute extraction 821 is used to extract attribute information from the fingerprint image; algorithm selection 822 is used to obtain multiple feature matching algorithms that match the attribute information output by attribute extraction 821; and fingerprint features are extracted from the fingerprint image according to these multiple feature matching algorithms to obtain multiple fingerprint features; feature fusion 823 is used to fuse the multiple fingerprint features output by algorithm selection 822 to obtain the target feature.

[0126] The difference between the registration branch and the retrieval branch is that the target feature obtained by the registration branch is the fingerprint sample feature, while the target feature obtained by the retrieval branch is the fingerprint feature to be identified.

[0127] Register branch 83: After obtaining the fingerprint sample feature 831, the fingerprint sample feature 831 can be stored in the fingerprint database 832 to provide a basis and support for fingerprint recognition.

[0128] Search branch 84: After obtaining the fingerprint feature 841 to be identified, fingerprint recognition can be performed on the fingerprint in the obtained fingerprint image based on the fingerprint feature 841 to be identified and the fingerprint sample feature 831 stored in the fingerprint database 832.

[0129] Taking two-level matching as an example, the fingerprint recognition process is explained as follows: By executing step 842, the fingerprint feature 841 to be identified is segmented to obtain n target feature sub-segments, and the first target feature group is obtained from the first r target feature sub-segments, and the second target feature group is obtained from the last (nr) target feature sub-segments.

[0130] For each fingerprint sample feature 831 in the fingerprint database 832, step 843 is executed to segment the currently traversed fingerprint sample feature 831 to obtain n sample feature sub-segments. The first sample feature group is obtained from the first r sample feature sub-segments, and the second sample feature group is obtained from the last (nr) sample feature sub-segments.

[0131] For the two target feature groups and two sample feature groups obtained, step 844 is executed to calculate the similarity between the corresponding target feature groups and sample feature groups. Specifically, the similarity between the first target feature group and the first sample feature group is calculated and denoted as Sa; the similarity between the second target feature group and the second sample feature group is calculated and denoted as Sb.

[0132] By executing steps 845 to 846, it is determined whether the fingerprint recognition was successful. Specifically, assuming T is the similarity threshold for biometric recognition, if... If the first-level matching fails, it means that the fingerprint recognition for the first target feature group has failed. At this point, the process proceeds to the second-level matching, i.e., fingerprint recognition continues for the first two target feature groups. If the second-level matching is successful, it means that the fingerprint recognition for the first two target feature groups was successful, or it can be considered that the fingerprint recognition in the acquired fingerprint image was successful. and These refer to the weights assigned to the first r target / sample feature segments and the last (nr) target / sample feature segments, respectively.

[0133] In this application scenario, on the one hand, by adding algorithm selection 822 to the fingerprint feature extraction architecture, more effective feature matching algorithms can be adapted for fingerprint images with different attribute information, at a low cost. On the other hand, by adapting the feature matching algorithm, the problem of low fingerprint recognition success rate caused by the large differences in fingerprints among different groups of people and under different environments can be effectively solved, making this fingerprint recognition scheme highly applicable and robust. In addition, the feature segmentation matching strategy can also effectively improve the fingerprint recognition speed, thereby effectively improving the fingerprint recognition efficiency.

[0134] In another application scenario, the image acquisition device can connect to the user's corresponding terminal via a network. The user can configure the image acquisition device through an app on the terminal, including configuring basic functions such as fingerprint enrollment, password setting, and anomaly reporting. Furthermore, the user can also customize the image recognition algorithm.

[0135] Specifically, all feature matching algorithms in the matching algorithm library are made available to users. This means that the feature matching algorithms in the library can be displayed in the terminal app. During the configuration of the image acquisition device, users can customize and select at least two feature matching algorithms from the library. In addition to displaying the various feature matching algorithms in the library, the attribute information of each algorithm can also be shown. For example, the target object can be a fingerprint, and the attribute information can include fingerprint quality, fingerprint type, etc. Therefore, when making a selection, users can customize and select feature matching algorithms that match their own biometric attributes, such as fingerprint attributes.

[0136] Then, based on at least two feature matching algorithms selected by the user, the feature matching algorithm currently used by the image acquisition device for image recognition processing is constructed.

[0137] In the process of recognizing the acquired images, the image acquisition device first acquires the image to be recognized, then uses at least one feature matching algorithm based on user-defined selection to extract image features from the image to be recognized, and then fuses the multiple image features to obtain the target features; then, the target object in the image to be recognized is identified according to the target features to obtain the recognition result.

[0138] In this application scenario, a matching algorithm library is provided, encompassing a variety of diverse and relatively accurate feature matching algorithms to support more effective feature matching for images with varying fingerprint quality. Furthermore, multiple matching algorithms from the library can be displayed to the user, allowing them to customize their selection based on their own biometric attributes. This not only effectively ensures the accuracy of biometric recognition but also significantly improves its adaptability and flexibility.

[0139] The following are embodiments of the apparatus described in this application, which can be used to execute the image recognition processing method involved in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the method embodiments of the image recognition processing method involved in this application.

[0140] Please see Figure 11 This application provides an image recognition processing device 900, including but not limited to: an algorithm acquisition module 910, a feature extraction module 930, a feature fusion module 950, and a feature recognition module 970.

[0141] Among them, the algorithm acquisition module 910 is used to acquire the image to be recognized and acquire multiple feature matching algorithms, and the feature matching algorithms are adapted to the attribute information of the target object in the image to be recognized.

[0142] The feature extraction module 930 is used to extract image features from the image to be identified according to various feature matching algorithms, thereby obtaining multiple image features.

[0143] The feature fusion module 950 is used to fuse multiple image features to obtain the target features.

[0144] The feature recognition module 970 is used to identify target objects in the image to be recognized based on target features, and obtain recognition results.

[0145] In an exemplary embodiment, the algorithm acquisition module includes: an information extraction unit for extracting attribute information from an image to be identified; and an algorithm selection unit for selecting multiple feature matching algorithms that are compatible with the attribute information from an algorithm set, wherein the algorithm set includes multiple candidate algorithms for image feature extraction that are available for selection.

[0146] In an exemplary embodiment, the algorithm selection unit includes: an algorithm search subunit, used to search for various candidate algorithms in the algorithm set according to attribute information; and an algorithm adaptation subunit, used to select candidate algorithms whose adaptation degree meets the adaptation condition as feature matching algorithms based on the degree of adaptation between the candidate algorithms and the attribute information.

[0147] In an exemplary embodiment, the attribute information includes primary attributes and secondary attributes associated with the primary attributes; the apparatus further includes: a set construction module for constructing an algorithm set; the set construction module includes: an algorithm classification unit for classifying multiple candidate algorithms according to different primary attributes to obtain multiple algorithm categories, wherein the candidate algorithms in each algorithm category correspond to the same primary attribute; an algorithm adaptation unit for adapting the candidate algorithms in each algorithm category to the secondary attributes associated with their corresponding primary attributes; and a path construction unit for constructing a path between the adapted candidate algorithms and the secondary attributes, and configuring the adaptation degree for the constructed path to obtain the algorithm set.

[0148] In an exemplary embodiment, the algorithm search subunit includes a correspondence subunit, used to search for candidate algorithms that correspond to the secondary attributes in the algorithm set based on the correspondence between the secondary attributes and the candidate algorithms in the algorithm set.

[0149] In an exemplary embodiment, the feature recognition module includes: a feature segmentation unit for segmenting target features to obtain multiple target feature sub-segments; a feature recognition unit for performing biometric recognition on target objects in the image to be recognized according to each target feature group to obtain recognition results corresponding to each target feature group, wherein the target feature group includes a set number of target feature sub-segments from multiple target feature sub-segments; and a result generation unit for obtaining recognition results according to the recognition results corresponding to each target feature group.

[0150] In an exemplary embodiment, the feature recognition unit includes: a sample acquisition subunit, configured to acquire a sample feature group corresponding to each target feature group for each sample feature in the sample set, wherein the sample feature group includes a set number of sample feature segments from a plurality of sample feature segments, and the sample feature segments are obtained by segmentation processing of the sample features; and a similarity calculation subunit, configured to calculate the similarity between each target feature group and the acquired sample feature group, and use it as the recognition result corresponding to each target feature group.

[0151] In one exemplary embodiment, the target object includes a fingerprint.

[0152] It should be noted that the image recognition processing device provided in the above embodiments is only illustrated by the division of the above functional modules when performing biometric recognition. In actual applications, the above functions can be assigned to different functional modules as needed. That is, the internal structure of the image recognition processing device will be divided into different functional modules to complete all or part of the functions described above.

[0153] Furthermore, the image recognition processing apparatus and image recognition processing method embodiments provided in the above embodiments belong to the same concept, and the specific way in which each module performs operations has been described in detail in the method embodiments, and will not be repeated here.

[0154] Figure 12 A schematic diagram of the structure of an electronic device is shown according to an exemplary embodiment. This electronic device is suitable for... Figure 1 The image acquisition device 130, server 150, and user terminal 170 are shown in the implementation environment.

[0155] It should be noted that this electronic device is merely an example adapted to this application and should not be construed as providing any limitation on the scope of use of this application. Furthermore, this electronic device should not be interpreted as requiring or depending on any specific feature. Figure 12 One or more components of the exemplary electronic device 2000 shown.

[0156] The hardware structure of electronic devices 2000 can vary significantly due to differences in configuration or performance, such as... Figure 12 As shown, the electronic device 2000 includes: a power supply 210, an interface 230, at least one memory 250, and at least one central processing unit (CPU) 270.

[0157] Specifically, power supply 210 is used to provide operating voltage for various hardware devices on electronic device 2000.

[0158] Interface 230 includes at least one wired or wireless network interface 231 for interacting with external devices. For example, to perform... Figure 1 The diagram illustrates the interaction between gateway 110 and server 150 in the implementation environment.

[0159] Of course, in other examples adapted in this application, interface 230 may further include at least one serial-to-parallel conversion interface 233, at least one input / output interface 235, and at least one USB interface 237, etc. Figure 12 As shown, this does not constitute a specific limitation.

[0160] The memory 250 serves as a carrier for resource storage and can be a read-only memory, random access memory, disk, or optical disk, etc. The resources stored on it include the operating system 251, application programs 253, and data 255, etc., and the storage method can be temporary storage or permanent storage.

[0161] The operating system 251 is used to manage and control the various hardware devices and application programs 253 on the electronic device 2000, so as to enable the central processing unit 270 to perform calculations and processing on the massive data 255 in the memory 250. It can be Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0162] Application 253 is a computer program that performs at least one specific task on top of operating system 251, and may include at least one module ( Figure 12 (Not shown), each module may contain a computer program for the electronic device 2000. For example, the image recognition processing device may be considered as application program 253 deployed on the electronic device 2000.

[0163] Data 255 can be photos, pictures, etc. stored on a disk, or images to be identified, and is stored in memory 250.

[0164] The central processing unit 270 may include one or more processors and is configured to communicate with the memory 250 via at least one communication bus to read computer programs stored in the memory 250, thereby performing operations and processing on massive amounts of data 255 stored in the memory 250. For example, an image recognition processing method may be implemented by the central processing unit 270 reading a series of computer programs stored in the memory 250.

[0165] Furthermore, this application can also be implemented through hardware circuits or a combination of hardware circuits and software. Therefore, the implementation of this application is not limited to any specific hardware circuit, software, or combination thereof.

[0166] Please see Figure 13 This application provides an electronic device 4000, which may include: fingerprint lock, access control, smart door lock, gateway camera, smartphone, desktop computer, laptop computer, server, etc.

[0167] exist Figure 13 The electronic device 4000 includes at least one processor 4001, at least one communication bus 4002, and at least one memory 4003.

[0168] The processor 4001 and memory 4003 are connected, for example, via a communication bus 4002. Optionally, the electronic device 4000 may also include a transceiver 4004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of this application.

[0169] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 4001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0170] The communication bus 4002 may include a path for transmitting information between the aforementioned components. The communication bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus 4002 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 13 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0171] The memory 4003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0172] The memory 4003 stores a computer program, and the processor 4001 reads the computer program stored in the memory 4003 through the communication bus 4002.

[0173] When the computer program is executed by the processor 4001, it implements the image recognition processing methods in the above embodiments.

[0174] Furthermore, this application provides a storage medium storing a computer program, which, when executed by a processor, implements the image recognition processing methods described in the above embodiments.

[0175] This application provides a computer program product comprising a computer program stored in a storage medium. A processor of a computer device reads the computer program from the storage medium and executes the computer program, causing the computer device to perform the image recognition processing methods described in the above embodiments.

[0176] Compared with related technologies, on the one hand, by constructing an algorithm set, more effective feature matching algorithms can be adapted for fingerprint images with different attribute information, which is low-cost; on the other hand, by adapting the feature matching algorithm, more effective feature matching algorithms can be obtained for images to be identified with different fingerprint qualities, so that the biometric identification of the images to be identified using the feature matching algorithm has a higher success rate, making the biometric identification scheme more applicable and robust; in addition, the feature segmentation matching strategy can also effectively improve the identification speed of biometric identification, thereby effectively improving the identification efficiency of biometric identification.

[0177] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0178] The above description is only a partial embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for human biometric identification, characterized in that, The method includes: The system acquires attribute information of multiple categories of the image to be identified and the target object in the image to be identified, and acquires multiple feature matching algorithms for each category of attribute information. The feature matching algorithms are adapted to the attribute information of the target object in the image to be identified. The attribute information is used to indicate any one of the identifiability and type of the target object. The target object includes human biometric features. Based on various feature matching algorithms, image features are extracted from the image to be identified to obtain multiple image features; The target features are obtained by fusing multiple image features. The target object in the image to be identified is identified based on the target features, and the identification result is obtained.

2. The method as described in claim 1, characterized in that, The process involves acquiring attribute information for multiple categories of the image to be identified and the target objects within that image, and then acquiring multiple feature matching algorithms for each category of attribute information, including: Multiple categories of attribute information are extracted from the image to be identified; From the algorithm set, select multiple feature matching algorithms that are adapted to the attribute information of each category, the algorithm set including multiple candidate algorithms for image feature extraction.

3. The method as described in claim 2, characterized in that, The selection of multiple feature matching algorithms from the algorithm set that are adapted to the attribute information of each category includes: Based on the attribute information of each category, multiple candidate algorithms corresponding to the attribute information of each category are searched in the algorithm set; Based on the degree of fit between the candidate algorithm and the attribute information, the candidate algorithm whose degree of fit meets the fit condition is selected as the feature matching algorithm.

4. The method as described in claim 3, characterized in that, The attribute information includes primary attributes and secondary attributes associated with the primary attributes; The method further includes: constructing the algorithm set; The construction of the algorithm set includes: Multiple candidate algorithms are classified according to different primary attributes to obtain multiple algorithm categories. The candidate algorithms in each algorithm category correspond to the same primary attribute. For each candidate algorithm in each algorithm category, the candidate algorithm is adapted to the secondary attributes associated with its corresponding primary attribute; A path is constructed between the suitable candidate algorithms and the secondary attributes, and the degree of adaptation is configured for the constructed path to obtain the algorithm set.

5. The method as described in claim 4, characterized in that, The step of searching for multiple candidate algorithms corresponding to the attribute information of each category in the algorithm set based on the attribute information of each category includes: Based on the correspondence between the secondary attributes and candidate algorithms in the algorithm set, candidate algorithms that correspond to the secondary attributes are searched from the algorithm set.

6. The method as described in claim 1, characterized in that, The step of identifying the target object in the image to be identified based on the target features to obtain the identification result includes: The target features are segmented to obtain multiple target feature sub-segments; Biometric identification is performed on the target object in the image to be identified according to each target feature group to obtain the identification result corresponding to each target feature group. The target feature group includes a set number of target feature segments from multiple target feature segments. The recognition result is obtained based on the recognition result corresponding to each of the target feature groups.

7. The method as described in claim 6, characterized in that, The step of performing biometric identification on the target object in the image to be identified based on each target feature group to obtain the identification result corresponding to each target feature group includes: For each sample feature in the sample set, obtain the sample feature group corresponding to each target feature group. The sample feature group includes a set number of sample feature segments from multiple sample feature segments. The sample feature segments are obtained by segmenting the sample features. Calculate the similarity between each target feature group and the obtained sample feature group, and use it as the recognition result corresponding to each target feature group.

8. The method according to any one of claims 1 to 7, characterized in that, The target object includes fingerprints.

9. A human biometric identification device, characterized in that, The device includes: The algorithm acquisition module is used to acquire the image to be identified and the attribute information of the target object in the image to be identified in multiple categories, and to acquire multiple feature matching algorithms for each category of attribute information. The feature matching algorithms are adapted to the attribute information of the target object in the image to be identified. The attribute information is used to indicate any one of the identifiability and type of the target object. The target object includes human biometric features. The feature extraction module is used to extract image features from the image to be identified according to various feature matching algorithms to obtain multiple image features; The feature fusion module is used to fuse multiple image features to obtain target features; The feature recognition module is used to identify the target object in the image to be recognized based on the target features, and obtain the recognition result.

10. An electronic device, characterized in that, include: At least one processor, at least one memory, and at least one communication bus, wherein, The memory stores a computer program, and the processor reads the computer program from the memory via the communication bus; When the computer program is executed by the processor, it implements the human biometric identification method according to any one of claims 1 to 8.

11. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the human biometric identification method as described in any one of claims 1 to 8.

Images