Method and apparatus for identifying a dog

Abstract

This application provides a dog identification method and apparatus. The dog identification method includes: acquiring at least two fingerprint images corresponding to a dog to be identified, wherein the fingerprint images to be identified are fingerprint images of palm prints from at least two acquisition angles corresponding to the dog to be identified; extracting features from each fingerprint image to obtain feature representations of at least two fingerprint images to be identified; searching in a preset feature representation set based on each fingerprint image feature representation to obtain a set of candidate dog identifiers corresponding to the dog to be identified; and determining the target dog identifier of the dog to be identified based on the candidate image set corresponding to the candidate dog identifier set and the at least two fingerprint images to be identified. Accurate identification of dogs is achieved through multi-angle dog palm prints, which not only improves the efficiency of dog identification but also enhances accuracy and user experience.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method for dog identification. This application also relates to a dog identification device, a computing device, and a computer-readable storage medium. Background Technology

[0002] As people's living standards improve, more and more people are keeping dogs, making better management of dogs increasingly important. Currently, there are limited technological means for dog identification, mainly relying on chip injection, nose print recognition, and facial recognition. Each method has its advantages and disadvantages. For example, chip injection can harm the dog's body and requires annual re-examination, which is expensive, time-consuming, and labor-intensive. Nose print and facial recognition methods are currently immature, lacking specific filtering of image quality and angles, resulting in low efficiency, low accuracy, and a poor user experience. Summary of the Invention

[0003] In view of this, embodiments of this application provide a dog identification method. This application also relates to a dog identification device, a computing device, and a computer-readable storage medium, to solve the aforementioned problems existing in the prior art.

[0004] According to a first aspect of the embodiments of this application, a dog identification method is provided, comprising:

[0005] Acquire at least two fingerprint images corresponding to the dog to be identified, wherein the fingerprint images are fingerprint images of the palm prints at at least two acquisition angles corresponding to the dog to be identified;

[0006] Feature extraction is performed on each fingerprint image to be identified to obtain at least two feature representations of the fingerprint images to be identified;

[0007] Based on the feature representation of each fingerprint image to be identified, a search is performed in a preset feature representation set to obtain a set of candidate dog identifiers corresponding to the dog to be identified.

[0008] Based on the candidate image set corresponding to the candidate dog identifier set and the at least two fingerprint images to be identified, the target dog identifier of the dog to be identified is determined.

[0009] According to a second aspect of the embodiments of this application, a dog identification device is provided, comprising:

[0010] The image acquisition module is configured to acquire at least two fingerprint images corresponding to the dog to be identified, wherein the fingerprint images to be identified are fingerprint images of the palm prints at at least two acquisition angles corresponding to the dog to be identified;

[0011] The feature extraction module is configured to extract features from each fingerprint image to be identified, and obtain at least two feature representations of the fingerprint images to be identified.

[0012] The vector retrieval module is configured to search within a preset feature representation set based on the feature representation of each fingerprint image to be identified, and obtain a set of candidate dog identifiers corresponding to the dog to be identified.

[0013] The dog identification module is configured to determine the target dog identifier of the dog to be identified based on the candidate image set corresponding to the candidate dog identifier set and the at least two fingerprint images to be identified.

[0014] According to a third aspect of the present application, a computing device is provided, including a memory, a processor, and computer instructions stored in the memory and executable on the processor, wherein the processor executes the computer instructions to implement the steps of the dog identification method.

[0015] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided that stores computer instructions which, when executed by a processor, implement the steps of the dog identification method.

[0016] The dog identification method provided in this application acquires at least two fingerprint images corresponding to a dog to be identified, wherein the fingerprint images to be identified are palm print images of at least two acquisition angles corresponding to the dog to be identified; performs feature extraction on each fingerprint image to obtain feature representations of at least two fingerprint images to be identified; searches in a preset feature representation set based on each fingerprint image feature representation to obtain a set of candidate dog identifiers corresponding to the dog to be identified; and determines the target dog identifier of the dog to be identified based on the candidate image set corresponding to the candidate dog identifier set and the at least two fingerprint images to be identified.

[0017] In one embodiment of this application, palm print images of a dog to be identified are acquired from different collection angles, and features are extracted from the palm print images to obtain a feature representation of the print image. Based on the feature representation of the print image, an initial screening is performed in a preset feature representation set to obtain a set of candidate dog identifiers. Then, in the set of candidate dog identifiers, the target dog identifier of the dog to be identified is selected again based on the matching of the print images to obtain the target identity information of the dog to be identified. This method achieves accurate identification of the dog's identity through multi-angle dog palm prints, which not only avoids the difficulties of image acquisition when identifying dog nose prints and dog faces, but also improves the efficiency of dog identification and enhances the user experience. Attached Figure Description

[0018] Figure 1This is a schematic flowchart of a dog identification method provided in an embodiment of this application;

[0019] Figure 2 This is a flowchart of a dog identification method provided in one embodiment of this application;

[0020] Figure 3 This is a schematic diagram of the structure of a dog identification device provided in one embodiment of this application;

[0021] Figure 4 This is a structural block diagram of a computing device provided in one embodiment of this application. Detailed Implementation

[0022] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0023] The terminology used in one or more embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the scope of one or more embodiments of this application. The singular forms “a,” “the,” and “the” used in one or more embodiments of this application and in the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” used in one or more embodiments of this application refers to and includes any or all possible combinations of one or more associated listed items.

[0024] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this application, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this application, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0025] First, the terms and concepts involved in one or more embodiments of this application will be explained.

[0026] ROI (Region of Interest): In machine vision and image processing, the region to be processed is delineated from the image using shapes such as rectangles, circles, ellipses, and irregular polygons.

[0027] CNN Architecture: A convolutional neural network architecture, including convolutional layers, downsampling layers, and fully connected layers. A convolutional neural network is a multi-layered supervised learning neural network. The hidden convolutional layers and pooling layers are the core modules for implementing the feature extraction function of the convolutional neural network. This network model uses gradient descent to minimize the loss function and adjusts the weight parameters in the network layer by layer inversely, improving the network's accuracy through frequent iterative training.

[0028] U-Net (Semantic Segmentation Network): Semantic segmentation requires determining the category of each pixel in an image for accurate segmentation.

[0029] Currently, traditional dog identification methods mostly rely on nose print recognition and face recognition. However, image acquisition for nose print and face recognition faces certain challenges, such as lighting conditions, focus issues, and water blurring, all of which affect image acquisition quality and consequently, the accuracy of dog identification. Therefore, this application provides a multi-pose (angle) local feature matching method. First, the most complete outline of the print is selected as the localization anchor point based on morphological features. Then, prints from different poses of the paw pad are acquired as feature extensions. Simultaneously, based on a local feature matching model, the pose deviation is estimated by calculating the affine matrix corresponding to the print. An index for each texture instance is then established based on the pose deviation. During the identification stage, local feature matching is used to calculate the pose deviation between the query and each instance anchor point, querying the corresponding print image for that pose, thus achieving texture region localization. The localized query image is then processed by a feature extraction model to obtain global features, and the distance to the corresponding instance image is calculated as the final ranking score to identify the specific dog's identity.

[0030] This application provides a dog identification method, and also relates to a dog identification device, a computing device, and a computer-readable storage medium, which will be described in detail in the following embodiments.

[0031] Figure 1 A flowchart illustrating a dog identification method according to an embodiment of this application is shown.

[0032] It should be noted that this embodiment provides a method for identifying the identity information of a dog. Due to the special shape and texture structure of a dog's paw pads, by collecting the paw print image of the dog's paw pads, the local features of the image can be obtained more clearly, ensuring the accuracy of subsequent dog identification.

[0033] In practical applications, a fingerprint image of the dog to be identified can be obtained. This fingerprint image may include a target fingerprint image, candidate fingerprint image 1, and candidate fingerprint image 2. It should be noted that the target fingerprint image can be understood as an image that meets the characteristic conditions determined among multiple fingerprint images, such as the integrity of the fingerprint and the clarity of the fingerprint texture. The candidate fingerprint image can be understood as a fingerprint image with a different acquisition angle from the target fingerprint image, including but not limited to fingerprint images with different acquisition angles based on the pressure intensity and direction of the pressure between the palm print and the acquisition surface. In this embodiment, in addition to the target fingerprint image, the acquisition of two fingerprint images with different acquisition angles is used as an example for explanation.

[0034] Furthermore, image features are extracted from the target fingerprint image, candidate fingerprint image 1, and candidate fingerprint image 2 respectively to obtain the target fingerprint image feature representation, candidate fingerprint image feature representation 1, and candidate fingerprint image feature representation 2. It should be noted that the extraction of image feature representations can adopt a preset feature extraction model, which is not specifically limited in this embodiment. Since the palm print acquisition angle corresponding to each image vector is different, the image vector corresponding to each acquisition angle can be searched in the preset feature representation set to achieve initial screening based on the feature representation. For example, the target fingerprint image feature representation is matched in the preset feature representation set 1, the candidate fingerprint image feature representation 1 is matched in the preset feature representation set 2, and the candidate fingerprint image feature representation 2 is matched in the preset feature representation set 3. Each feature representation set pre-stores the palm print image feature representations of all dogs acquired at a certain acquisition angle. Therefore, in each search result, the corresponding candidate dog identifier can be determined, and a candidate dog identifier set can be generated.

[0035] Furthermore, after obtaining the candidate dog identifier, the candidate fingerprint image corresponding to the candidate dog identifier can be selected from the preset fingerprint image set. Then, according to the acquisition angle, the target fingerprint image, candidate fingerprint image 1, and candidate fingerprint image 2 are matched with the corresponding candidate fingerprint images to obtain the matching result of each pair of images. The matching results are sorted, and the target dog identifier of the dog to be identified is determined in the candidate dog identifier set, thus clarifying the specific identity information of the dog to be identified.

[0036] In summary, the dog identification method provided in this application collects dog paw print images corresponding to multiple collection angles of the dog to be identified, and uses the feature representation of the print image corresponding to each collection angle to initially screen out a set of candidate dog identifiers, and then further uses the dog paw print images to perform fine image matching in order to accurately obtain the target dog identifier of the dog to be identified.

[0037] Figure 2A flowchart of a dog identification method according to an embodiment of this application is shown, which specifically includes the following steps:

[0038] Step 202: Obtain at least two fingerprint images corresponding to the dog to be identified, wherein the fingerprint images are fingerprint images of the palm prints at at least two acquisition angles corresponding to the dog to be identified.

[0039] The dog to be identified can be understood as a dog whose identity needs to be identified; it should be noted that the identity information of the dog to be identified can be either already recorded in the preset dog database or not recorded, and this embodiment does not make a specific limitation on this.

[0040] The fingerprint image to be identified can be understood as the fingerprint image corresponding to the palm print of the dog to be identified, and the identity of the dog to be identified is identified using this fingerprint image.

[0041] In practical applications, at least two fingerprint images of the dog to be identified can be obtained first. These at least two fingerprint images are palm print images corresponding to different angles collected from the dog to be identified. The collection angle can be understood as the angle between the pressing angle of a certain paw of the dog to be identified and the collection contact surface, or as the angle of the circumference of the edge connecting the paw and the collection contact surface. In this embodiment, no specific limitation is made on this. The palm print images corresponding to different collection angles are not the same.

[0042] Step 204: Extract features from each fingerprint image to be identified, and obtain feature representations of at least two fingerprint images to be identified.

[0043] Furthermore, after acquiring multiple fingerprint images corresponding to the dog to be identified, feature extraction can be performed on each fingerprint image to obtain the image feature representation corresponding to each fingerprint image, so as to facilitate feature matching based on the image feature representation corresponding to each acquisition angle, thereby realizing the fingerprint image recognition process.

[0044] Specifically, feature extraction is performed on each fingerprint image to be identified, obtaining at least two feature representations of the fingerprint image to be identified, including:

[0045] Based on a preset image segmentation model, features are extracted from each print image to be identified, and the feature representation of the print image to be identified corresponding to each acquisition angle is obtained.

[0046] Among them, the preset image segmentation model can be understood as a model that segments information in an image, such as the U-Net model and other semantic segmentation models.

[0047] In practical applications, the feature extraction process can be implemented using a pre-defined image segmentation model, such as the U-Net model, to extract the effective region of the palm print in the image to be identified and filter out the interference region of the palm print. Because when multiple palm print images of the dog to be identified are acquired by the image acquisition device, there may be interference content in each image, so the image segmentation component needs to filter out these interference contents (such as body hair) to extract the effective region of the palm print. Then, feature extraction is performed on the image in the effective region to obtain the feature representation of the image to be identified. The feature representation of the image to be identified can represent the information of each feature point in the image. It should be noted that the method of filtering the effective region can not only use image semantic segmentation technology, but also other effective region extraction techniques, such as image matting, etc., without specific limitations.

[0048] Step 206: Based on the feature representation of each fingerprint image to be identified, search in the preset feature representation set to obtain the candidate dog identifier set corresponding to the dog to be identified.

[0049] The preset feature representation set can be understood as the set of feature representations of palm print images corresponding to various collection angles of all dogs in the dog database. In other words, the preset feature representation set is a pre-constructed set of palm print image feature representations corresponding to all dogs in the database, which facilitates the subsequent use of the preset feature representation set to compare with the palm print image feature representation of the dog to be identified.

[0050] In practical applications, after obtaining the feature representation of the fingerprint image to be identified corresponding to each acquisition angle, the feature representation of the fingerprint image to be identified is retrieved in the preset feature representation set. This realizes the process of first retrieving based on the image feature representation, and initially obtains a set of candidate dog identifiers that are similar to the features of the fingerprint image feature representation of the dog to be identified, thus achieving the purpose of initial screening and identification based on the fingerprint image feature representation.

[0051] Furthermore, based on the feature representation of each fingerprint image to be identified, a search is performed in a preset feature representation set to obtain a set of candidate dog identifiers corresponding to the dog to be identified, including:

[0052] Based on the acquisition angle corresponding to the feature representation of the fingerprint image to be identified, a preset feature representation subset corresponding to each acquisition angle is determined, wherein the preset feature representation subset is a set of pre-constructed feature representations of dog fingerprint images for each acquisition angle;

[0053] Based on the feature representation of the fingerprint image to be identified, matching is performed in the preset feature representation subset to obtain the candidate dog identifier corresponding to the preset feature representation subset;

[0054] Based on the candidate dog identifiers corresponding to the preset feature representation subset, a set of candidate dog identifiers corresponding to the dog to be identified is generated.

[0055] The preset feature representation subset can be understood as a subset of the feature representation of the preset feature representation set. Each feature representation subset is a set of fingerprint image vectors corresponding to the same acquisition angle for all dogs entering the warehouse. For example, the preset feature representation set includes a preset feature representation subset with an acquisition angle of 15 degrees and a preset feature representation subset with an acquisition angle of -15 degrees (the acquisition angle can be understood as the angle between the direction of fingerprint acquisition and the skew angle of the contact plane).

[0056] In practical applications, based on the acquisition angle corresponding to each fingerprint image feature representation to be identified, a preset feature representation subset corresponding to each acquisition angle is determined. This preset feature representation subset is a pre-constructed set of all dog fingerprint image feature representations for each acquisition angle. Then, each fingerprint image feature representation to be identified is searched in the corresponding preset feature representation subset. Each search yields the image feature representation with the most similar features, and the dog identifier corresponding to this image feature representation is used as the candidate dog identifier corresponding to the preset feature representation subset. Furthermore, the candidate dog identifiers selected from each preset feature representation subset can form a set of candidate dog identifiers corresponding to the dog to be identified, so that the target dog identifier can be further selected from the set of candidate dog identifiers to determine the identity of the dog to be identified.

[0057] It should be noted that a preset feature representation subset corresponding to different acquisition angles is established. That is, for the fingerprint acquisition process of each dog, the corresponding fingerprint image features are extracted according to the set acquisition angle. Then, the global features of the fingerprint images corresponding to different acquisition angles are extracted using a feature extraction model based on CNN architecture, and added to the feature representation subset of the corresponding acquisition angle respectively.

[0058] Step 208: Based on the candidate image set corresponding to the candidate dog identifier set and the at least two fingerprint images to be identified, determine the target dog identifier of the dog to be identified.

[0059] In practical applications, after determining the set of candidate dog identifiers, the candidate fingerprint image corresponding to each candidate dog identifier can be determined. The candidate fingerprint images form a candidate image set. Based on the candidate image set and at least two fingerprint images to be identified, the target dog identifier of the dog to be identified can be determined, that is, the identity information of the dog to be identified can be obtained.

[0060] Furthermore, after determining the candidate dog identifier set, the screening process for identifying the dog to be identified has been narrowed down. Subsequently, the target dog identifier can be precisely determined from the candidate dog identifier set, thereby obtaining the identification information of the dog to be identified. Specifically, based on the candidate image set corresponding to the candidate dog identifier set and the at least two fingerprint images to be identified, the target dog identifier of the dog to be identified is determined, including:

[0061] Based on the candidate dog identifier set in the preset image set, a candidate image set is determined, wherein the candidate image set includes a candidate fingerprint image corresponding to each candidate dog;

[0062] Based on the candidate image set and the at least two fingerprint images to be identified, image matching is performed to determine the target dog identifier of the dog to be identified.

[0063] In practical applications, based on each candidate dog identifier in the candidate dog identifier set, a candidate fingerprint image corresponding to each candidate dog identifier is determined in a preset image set. Based on the candidate fingerprint image corresponding to each candidate dog identifier and at least two fingerprint images to be identified, image matching is performed to determine the target dog identifier of the dog to be identified.

[0064] In practice, the preset image set is a pre-constructed set of dog print images, and the specific construction steps are as follows:

[0065] Determine the target image of the target palm print acquisition angle corresponding to the dog to be processed;

[0066] Acquire at least one reference image of the dog to be processed that has a different acquisition angle from the target palm print;

[0067] The target image and each reference image are matched for feature points to determine the image deviation angle corresponding to each reference image.

[0068] A preset image set is constructed based on the target image and the target palm print acquisition angle, the at least one reference image and the image deviation angle corresponding to each reference image.

[0069] In practical applications, the preset image set is a collection of palm print images from various acquisition angles corresponding to all dogs in the database. This image set is a set of <palm print images, angles>. Specifically, the dog to be processed can be understood as each dog in the database. Correspondingly, a target image corresponding to the target palm print acquisition angle of the dog to be processed is acquired. This target image can be understood as the image of the dog to be processed at the target acquisition angle. The target acquisition angle can be the angle at which the plane of the palm print makes positive contact with the acquisition contact surface, which can clearly obtain the angle of the overall palm print of the dog to be processed. Furthermore, other reference images with different acquisition angles from the target palm print can also be acquired. These can be understood as palm print images with a skewed angle at which the plane of the palm print makes positive contact with the acquisition contact surface. Then, feature point matching is performed between each reference image and the target image to obtain an affine matrix of image transformation. The image deviation angle of each reference image relative to the target image is then determined based on the affine matrix, such as <reference image 1, 15°>. Finally, the target image and the target palm print acquisition angle, as well as multiple reference images and their corresponding acquisition angles, are used to construct the preset image set.

[0070] It should be noted that the reference image can be understood as the imprint image obtained by the collector by changing the direction of the pressure and the orientation of the contact surface at different angles. Then, the affine matrix between the reference image and the target image is estimated using a local feature matching algorithm to obtain the pose deviation and label the current reference image. Simultaneously, a threshold is set based on the manually defined proportion of effectively matched feature point regions, and the reference image is added to the feature imprint set (preset image set). Finally, a set of images with pairs of <imprint image, angle> is obtained. Local feature matching: First, imprint feature points and descriptors are extracted based on SIFT. Then, MNN+RANSAC matching is used to calculate the matching relationship between the imprint and anchor point image feature points, and the affine matrix is ​​obtained by minimizing the projection error, which serves as the image deviation angle for this image. This embodiment does not specifically limit this.

[0071] Furthermore, the target image corresponding to the target palm print acquisition angle of the dog to be processed is determined, including:

[0072] Based on a preset image acquisition device, multiple initial image frames of the paw print of the dog to be processed are acquired.

[0073] Based on preset image filtering conditions, the target image is determined by filtering among the multiple initial image frames, and the target palmprint acquisition angle corresponding to the target image is determined.

[0074] In specific implementation, multiple initial image frames of the paw print of the dog to be processed are acquired by a preset image acquisition device. Then, according to preset image filtering conditions, the target image and the target paw print acquisition angle corresponding to the target image are determined by filtering among the multiple initial image frames. The preset image acquisition device can be any device that can accurately acquire continuous frames, such as an optical sensor. This embodiment does not make specific limitations on this.

[0075] In practical applications, optical sensors can be used to acquire continuous initial image frames, and images that meet the criteria can be selected as target images from the continuous initial image frames. The preset image selection criteria include dog print symmetry selection criteria, dog print integrity selection criteria, and dog print texture intensity selection criteria.

[0076] In an optional embodiment, the symmetry of dog prints can be calculated by the difference in histogram distribution on both sides of the ROI symmetry axis; the integrity of the dog print shape can be calculated by comparing the Hu invariant moments of the ROI contour and the hand-constructed paw pad shape template; and the texture intensity of the dog print can be calculated based on the contrast of the gray-level co-occurrence matrix.

[0077] When performing image matching on the fingerprint image to be identified based on local features, the target dog identifier corresponding to the fingerprint image to be identified is determined by matching fingerprint images under the same acquisition angle; specifically, image matching is performed based on the candidate image set and the at least two fingerprint images to be identified to determine the target dog identifier of the dog to be identified, including:

[0078] Determine the candidate pattern acquisition angle corresponding to each candidate pattern image in the candidate image set;

[0079] Determine the acquisition angle of the fingerprint image to be identified for each fingerprint image;

[0080] Based on the acquisition angle of the candidate fingerprint and the acquisition angle of the fingerprint to be identified, image matching is performed to determine the target dog identifier of the dog to be identified.

[0081] In practical applications, the candidate fingerprint acquisition angle corresponding to each candidate fingerprint image in the candidate image set is determined, the fingerprint acquisition angle corresponding to the fingerprint image to be identified is determined, and the matching process of the corresponding images is completed based on these two acquisition angles, so as to determine the target dog identifier of the dog to be identified based on the matching results.

[0082] Specifically, based on the acquisition angle of the candidate fingerprint and the acquisition angle of the fingerprint to be identified, image matching is performed to determine the target dog identifier of the dog to be identified, including:

[0083] Select the acquisition angle to be matched from the candidate pattern acquisition angles, and determine the candidate pattern image corresponding to the acquisition angle to be matched;

[0084] Select an identification acquisition angle that is the same as the matching acquisition angle from the acquisition angles of the fingerprint to be identified, and determine the fingerprint image to be identified corresponding to the acquisition angle;

[0085] The candidate pattern image and the pattern image to be identified are matched to determine the matching score corresponding to the acquisition angle to be identified.

[0086] Based on the matching score, the target dog identifier of the dog to be identified is determined.

[0087] In practical applications, each image matching process requires first determining the acquisition angle to be matched, and then obtaining the corresponding candidate fingerprint image and the fingerprint image to be identified based on the acquisition angle. In this way, the fingerprint images under each acquisition angle are matched to determine the matching score corresponding to each acquisition angle to be identified. Finally, the target dog identifier of the dog to be identified is determined based on multiple matching scores.

[0088] As an example, the candidate image set contains only three candidate fingerprint images with acquisition angles of -15°, 0°, and 15°. The fingerprint image to be identified also contains fingerprint images with the above three acquisition angles. Therefore, image matching is performed according to the acquisition angle. For example, the candidate fingerprint image with an acquisition angle of 0° is matched with the fingerprint image to be identified with an acquisition angle of 0°, the candidate fingerprint image with an acquisition angle of 15° is matched with the fingerprint image to be identified with an acquisition angle of 15°, and the candidate fingerprint image with an acquisition angle of -15° is matched with the fingerprint image to be identified with an acquisition angle of -15°. Based on the three matching results, the target dog identifier of the dog to be identified is determined.

[0089] In addition, this embodiment can also provide another image matching method. For example, the fingerprint image to be identified with a collection angle of 0° is matched with candidate fingerprint images with three collection angles to obtain three matching sub-results. Then, a matching score of 1 is calculated based on these three matching sub-results. The fingerprint image to be identified with a collection angle of 15° is matched with candidate fingerprint images with three collection angles to obtain three matching sub-results. Then, a matching score of 2 is calculated based on these three matching sub-results. The fingerprint image to be identified with a collection angle of -15° is matched with candidate fingerprint images with three collection angles to obtain three matching sub-results. Then, a matching score of 3 is calculated based on these three matching sub-results. Finally, the target dog identifier of the dog to be identified is determined based on the three matching scores.

[0090] Therefore, this embodiment does not limit the above two image matching methods. It mainly uses the acquisition angle to classify and match the images to be matched in order to obtain the corresponding matching score, and then determines the target dog identifier of the dog to be identified based on the matching score.

[0091] Furthermore, after obtaining the matching degree of the fingerprint image at each acquisition angle, the ranking features corresponding to different acquisition angles can be obtained. Then, based on these ranking features, the fingerprint image with the highest matching degree is determined, and the dog identifier corresponding to this fingerprint image is the target dog identifier for the dog to be identified. Specifically, based on the matching score, determining the target dog identifier for the dog to be identified includes:

[0092] Determine the matching score corresponding to each angle to be identified;

[0093] Each matching score is input into the score sorting model to obtain the target matching score and the target dog identifier corresponding to the target matching score.

[0094] The score ranking model can be understood as a model that sorts matching scores according to preset rules. The specific ranking rules can be set differently according to different application scenarios, and this embodiment does not make specific limitations on this.

[0095] In practical applications, after determining the matching score of the image at each acquisition angle, the target dog identifier of the dog to be identified is determined based on the matching score. Specifically, each matching score can be input into the score ranking model to obtain the target matching score output by the score ranking model, as well as the target dog identifier corresponding to the target matching score. It should be noted that the target matching score can be understood as the highest matching score in the score ranking. Therefore, the fingerprint image corresponding to the target matching score can be understood as the fingerprint image corresponding to the dog to be identified. Consequently, the dog identifier corresponding to the fingerprint image is the target dog identifier corresponding to the dog to be identified, thus obtaining the specific identity of the dog to be identified.

[0096] In summary, the dog identification method provided in this application acquires palm print images of the dog to be identified from various acquisition angles. Then, for each palm print image, it searches in the feature vector index table (a pre-constructed set of palm print image feature representations) for its corresponding acquisition angle, merges the ranking results of vector matching corresponding to different angles, and obtains a top-k identification ranking (ranking in the candidate dog identifier set). Based on the obtained top-k preliminary ranking results, local feature matching is performed for each palm print image. That is, the matching process calculates the matching degree between the query image and the instance image (a pre-constructed set of palm print images) based on the same acquisition angle, obtains the corresponding ranking features (such as matching scores) under different angles, and outputs them to the ranking model to obtain the final matching score. Then, based on the matching score, the target dog identifier corresponding to the dog to be identified can be accurately output.

[0097] Corresponding to the above method embodiments, this application also provides embodiments of a dog identification device. Figure 3 A schematic diagram of a dog identification device according to an embodiment of this application is shown. Figure 3 As shown, the device includes:

[0098] The image acquisition module 302 is configured to acquire at least two fingerprint images corresponding to the dog to be identified, wherein the fingerprint images to be identified are fingerprint images of palm prints from at least two acquisition angles corresponding to the dog to be identified.

[0099] The feature extraction module 304 is configured to extract features from each fingerprint image to be identified, and obtain at least two feature representations of the fingerprint images to be identified.

[0100] The vector retrieval module 306 is configured to search in a preset feature representation set based on the feature representation of each fingerprint image to be identified, and obtain a set of candidate dog identifiers corresponding to the dog to be identified.

[0101] The dog identification module 308 is configured to determine the target dog identifier of the dog to be identified based on the candidate image set corresponding to the candidate dog identifier set and the at least two fingerprint images to be identified.

[0102] Optionally, the dog identification module 308 is further configured to:

[0103] Based on the candidate dog identifier set in the preset image set, a candidate image set is determined, wherein the candidate image set includes a candidate fingerprint image corresponding to each candidate dog;

[0104] Based on the candidate image set and the at least two fingerprint images to be identified, image matching is performed to determine the target dog identifier of the dog to be identified.

[0105] Optionally, the preset image set is a pre-constructed set of dog print images.

[0106] Optionally, the device further includes a collection building module configured to:

[0107] Determine the target image of the target palm print acquisition angle corresponding to the dog to be processed;

[0108] Acquire at least one reference image of the dog to be processed that has a different acquisition angle from the target palm print;

[0109] The target image and each reference image are matched for feature points to determine the image deviation angle corresponding to each reference image.

[0110] A preset image set is constructed based on the target image and the target palm print acquisition angle, the at least one reference image and the image deviation angle corresponding to each reference image.

[0111] Optionally, the collection building module is further configured to:

[0112] Based on a preset image acquisition device, multiple initial image frames of the paw print of the dog to be processed are acquired.

[0113] Based on preset image filtering conditions, the target image is determined by filtering among the multiple initial image frames, and the target palmprint acquisition angle corresponding to the target image is determined.

[0114] Optionally, the preset image filtering conditions include dog print symmetry filtering conditions, dog print integrity filtering conditions, and dog print texture intensity filtering conditions.

[0115] Optionally, the dog identification module 308 is further configured to:

[0116] Determine the candidate pattern acquisition angle corresponding to each candidate pattern image in the candidate image set;

[0117] Determine the acquisition angle of the fingerprint image to be identified for each fingerprint image;

[0118] Based on the acquisition angle of the candidate fingerprint and the acquisition angle of the fingerprint to be identified, image matching is performed to determine the target dog identifier of the dog to be identified.

[0119] Optionally, the dog identification module 308 is further configured to:

[0120] Select the acquisition angle to be matched from the candidate pattern acquisition angles, and determine the candidate pattern image corresponding to the acquisition angle to be matched;

[0121] Select an identification acquisition angle that is the same as the matching acquisition angle from the acquisition angles of the fingerprint to be identified, and determine the fingerprint image to be identified corresponding to the acquisition angle;

[0122] The candidate pattern image and the pattern image to be identified are matched to determine the matching score corresponding to the acquisition angle to be identified.

[0123] Based on the matching score, the target dog identifier of the dog to be identified is determined.

[0124] Optionally, the dog identification module 308 is further configured to:

[0125] Determine the matching score corresponding to each angle to be identified;

[0126] Each matching score is input into the score sorting model to obtain the target matching score and the target dog identifier corresponding to the target matching score.

[0127] Optionally, the feature extraction module 304 is further configured to:

[0128] Based on a preset image segmentation model, features are extracted from each print image to be identified, and the feature representation of the print image to be identified corresponding to each acquisition angle is obtained.

[0129] Optionally, the vector retrieval module 306 is further configured to:

[0130] Based on the acquisition angle corresponding to the feature representation of the fingerprint image to be identified, a preset feature representation subset corresponding to each acquisition angle is determined, wherein the preset feature representation subset is a set of pre-constructed feature representations of dog fingerprint images for each acquisition angle;

[0131] Based on the feature representation of the fingerprint image to be identified, matching is performed in the preset feature representation subset to obtain the candidate dog identifier corresponding to the preset feature representation subset;

[0132] Based on the candidate dog identifiers corresponding to the preset feature representation subset, a set of candidate dog identifiers corresponding to the dog to be identified is generated.

[0133] The dog identification device provided in this application acquires palm print images of the dog to be identified from different acquisition angles, extracts features from the palm print images to obtain a feature representation of the print image, performs initial screening in a preset feature representation set based on the print image feature representation to obtain a set of candidate dog identifiers, and then, based on the matching of print images, further filters out the target dog identifier of the dog to be identified to obtain the target identity information of the dog to be identified. This device achieves accurate identification of dogs through multi-angle dog palm prints, which not only avoids the difficulties of image acquisition when identifying dog nose prints and dog faces, but also improves the efficiency of dog identification and enhances the user experience.

[0134] The above is a schematic scheme of a dog identification device according to this embodiment. It should be noted that the technical solution of this dog identification device and the technical solution of the above-described dog identification method belong to the same concept. For details not described in detail in the technical solution of the dog identification device, please refer to the description of the technical solution of the above-described dog identification method.

[0135] Figure 4 A structural block diagram of a computing device 400 according to an embodiment of this application is shown. The components of the computing device 400 include, but are not limited to, a memory 410 and a processor 420. The processor 420 is connected to the memory 410 via a bus 430, and a database 450 is used to store data.

[0136] The computing device 400 also includes an access device 440, which enables the computing device 400 to communicate via one or more networks 460. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 440 may include one or more of any type of wired or wireless network interface (e.g., a network interface controller (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, a Near Field Communication (NFC) interface, and so on.

[0137] In one embodiment of this application, the aforementioned components of the computing device 400 and Figure 4 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 4 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this application. Those skilled in the art can add or replace other components as needed.

[0138] Computing device 400 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). Computing device 400 can also be a mobile or stationary server.

[0139] The processor 420 executes the computer instructions to implement the steps of the dog identification method.

[0140] The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device and the technical solution of the above-described dog recognition method belong to the same concept. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the above-described dog recognition method.

[0141] An embodiment of this application also provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the steps of the dog identification method as described above.

[0142] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium and the technical solution of the above-described dog identification method belong to the same concept. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the above-described dog identification method.

[0143] The foregoing has described specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0144] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added to or subtracted according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0145] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0146] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0147] The preferred embodiments disclosed above are merely illustrative of this application. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this application. These embodiments are selected and specifically described in this application to better explain the principles and practical applications of this application, thereby enabling those skilled in the art to better understand and utilize this application. This application is limited only by the claims and their full scope and equivalents.

Claims

1. A method for dog identification, characterized in that, include: Acquire at least two fingerprint images corresponding to the dog to be identified, wherein the fingerprint images to be identified are fingerprint images of the palm prints of the dog to be identified at least two acquisition angles, and the acquisition angle is the angle between the pressing angle of a certain paw of the dog to be identified and the acquisition contact surface. Feature extraction is performed on each fingerprint image to be identified to obtain at least two feature representations of the fingerprint images to be identified; Based on the feature representation of each fingerprint image to be identified, a search is performed in a preset feature representation set to obtain a set of candidate dog identifiers corresponding to the dog to be identified. Based on the candidate dog identifier set in a preset image set, a candidate image set is determined. The candidate image set includes candidate fingerprint images corresponding to each candidate dog. The preset image set is a pre-constructed set of dog fingerprint images. The specific construction steps include: determining a target image with a target palm print acquisition angle corresponding to the dog to be processed; acquiring at least one reference image corresponding to the dog to be processed that has a different acquisition angle from the target palm print; performing feature point matching between the target image and each reference image to determine the image deviation angle corresponding to each reference image; constructing a preset image set based on the target image and the target palm print acquisition angle, the at least one reference image, and the image deviation angle corresponding to each reference image; and performing image matching based on the candidate image set and the at least two fingerprint images to be identified to determine the target dog identifier of the dog to be identified. The target image is the angle at which the plane of the palm print acquired by the dog to be processed makes positive contact with the acquisition contact surface, and the reference image is a palm print image with a skewed angle between the plane of the palm print and the acquisition contact surface.

2. The method as described in claim 1, characterized in that, The target image corresponding to the target paw print acquisition angle of the dog to be processed includes: Based on a preset image acquisition device, multiple initial image frames of the paw print of the dog to be processed are acquired. Based on preset image filtering conditions, the target image is determined by filtering among the multiple initial image frames, and the target palmprint acquisition angle corresponding to the target image is determined.

3. The method as described in claim 1, characterized in that, The preset image filtering conditions include dog print symmetry filtering conditions, dog print integrity filtering conditions, and dog print texture intensity filtering conditions.

4. The method as described in claim 1, characterized in that, Based on the candidate image set and the at least two fingerprint images to be identified, image matching is performed to determine the target dog identifier of the dog to be identified, including: Determine the candidate pattern acquisition angle corresponding to each candidate pattern image in the candidate image set; Determine the acquisition angle of the fingerprint image to be identified for each fingerprint image; Based on the acquisition angle of the candidate fingerprint and the acquisition angle of the fingerprint to be identified, image matching is performed to determine the target dog identifier of the dog to be identified.

5. The method as described in claim 4, characterized in that, Based on the candidate fingerprint acquisition angle and the fingerprint acquisition angle to be identified, image matching is performed to determine the target dog identifier of the dog to be identified, including: Select the acquisition angle to be matched from the candidate pattern acquisition angles, and determine the candidate pattern image corresponding to the acquisition angle to be matched; Select an identification acquisition angle that is the same as the matching acquisition angle from the acquisition angles of the fingerprint to be identified, and determine the fingerprint image to be identified corresponding to the acquisition angle; The candidate pattern image and the pattern image to be identified are matched to determine the matching score corresponding to the acquisition angle to be identified. Based on the matching score, the target dog identifier of the dog to be identified is determined.

6. The method as described in claim 5, characterized in that, Based on the matching score, the target dog identifier of the dog to be identified is determined, including: Determine the matching score corresponding to each angle to be identified; Each matching score is input into the score sorting model to obtain the target matching score and the target dog identifier corresponding to the target matching score.

7. The method according to any one of claims 1-6, characterized in that, For each fingerprint image to be identified, feature extraction is performed to obtain at least two feature representations of the fingerprint image to be identified, including: Based on a preset image segmentation model, features are extracted from each print image to be identified, and the feature representation of the print image to be identified corresponding to each acquisition angle is obtained.

8. The method as described in claim 7, characterized in that, Based on the feature representation of each fingerprint image to be identified, a search is performed in a preset feature representation set to obtain a set of candidate dog identifiers corresponding to the dog to be identified, including: Based on the acquisition angle corresponding to the feature representation of the fingerprint image to be identified, a preset feature representation subset corresponding to each acquisition angle is determined, wherein the preset feature representation subset is a set of pre-constructed feature representations of dog fingerprint images for each acquisition angle; Based on the feature representation of the print image to be identified, a search is performed in the preset feature representation subset to obtain the candidate dog identifiers corresponding to the preset feature representation subset; Based on the candidate dog identifiers corresponding to the preset feature representation subset, a set of candidate dog identifiers corresponding to the dog to be identified is generated.

9. A dog identification device, characterized in that, include: The image acquisition module is configured to acquire at least two fingerprint images corresponding to the dog to be identified, wherein the fingerprint images to be identified are fingerprint images of the palm prints at at least two acquisition angles corresponding to the dog to be identified, and the acquisition angle is the angle between the pressing angle of a certain paw of the dog to be identified and the acquisition contact surface. The feature extraction module is configured to extract features from each fingerprint image to be identified, and obtain at least two feature representations of the fingerprint images to be identified. The vector retrieval module is configured to search within a preset feature representation set based on the feature representation of each fingerprint image to be identified, and obtain a set of candidate dog identifiers corresponding to the dog to be identified. A dog identification module is configured to determine a candidate image set based on a set of candidate dog identifiers within a preset image set. The candidate image set includes candidate fingerprint images corresponding to each candidate dog. The preset image set is a pre-constructed set of dog fingerprint images. The specific construction steps include: determining a target image with a target palm print acquisition angle corresponding to the dog to be processed; acquiring at least one reference image corresponding to the dog to be processed that has a different acquisition angle from the target palm print; performing feature point matching between the target image and each reference image to determine the image deviation angle corresponding to each reference image; constructing a preset image set based on the target image and the target palm print acquisition angle, and the at least one reference image and the image deviation angle corresponding to each reference image; and performing image matching based on the candidate image set and the at least two fingerprint images to be identified to determine the target dog identifier of the dog to be identified. The target image is an image showing the forward contact angle between the plane of the palm print acquired by the dog to be processed and the acquisition contact surface, and the reference images are palm print images with a skewed forward contact angle between the plane of the palm print and the acquisition contact surface.

10. A computing device, comprising a memory, a processor, and computer instructions stored in the memory and executable on the processor, characterized in that, When the processor executes the computer instructions, it implements the steps of the method according to any one of claims 1-8.

11. A computer-readable storage medium storing computer instructions, characterized in that, When executed by a processor, the computer instructions implement the steps of the method according to any one of claims 1-8.

Images