Image selection method, computer device and storage apparatus

Abstract

The application discloses an image selection method, a computer device and a storage device. The method comprises the following steps: obtaining a target image sequence of a target; classifying each target image in the target image sequence according to feature information of each target image, so as to obtain at least one classified image set of the target, and each classified image set comprises target images of one category; performing clustering processing on the at least one classified image set of the target, so as to obtain feature clustering data; and selecting a preferred target image corresponding to the target based on the feature clustering data corresponding to the target. The above scheme can improve the diversity of target image selection.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to an image selection method, computer equipment, and storage device. Background Technology

[0002] In recent years, with the continuous development of information technology, a large number of images are acquired in various application scenarios, such as video or image frames captured by video surveillance systems, images taken by smart terminals, and text images collected in various fields. Due to the large amount of data in the images, the storage pressure on the images is high.

[0003] Currently, one or more images are typically selected from multiple images of a target and stored based on their image quality. While this reduces the amount of images stored, the limitations of the image selection method can lead to overly homogeneous images being selected, resulting in poor actual usability of the images in the future. Summary of the Invention

[0004] The main technical problem addressed by this application is to provide an image selection method, computer device, and storage device that can improve the diversity of target image selection.

[0005] To address the aforementioned problems, the first aspect of this application provides an image selection method, comprising: acquiring a target image sequence of a target; classifying each target image according to feature information of each target image in the target image sequence to obtain at least one set of classified images of the target, each set of classified images containing target images of one category; performing clustering processing on the at least one set of classified images of the target to obtain feature clustering data; and selecting a preferred target image corresponding to the target based on the feature clustering data corresponding to the target.

[0006] To address the aforementioned problems, a second aspect of this application provides a computer device comprising a memory and a processor coupled to each other, wherein the memory stores program data and the processor executes the program data to implement any step of the aforementioned image selection method.

[0007] To address the aforementioned problems, a third aspect of this application provides a storage device that stores program data executable by a processor, the program data being used to implement any step of the aforementioned image selection method.

[0008] The above scheme obtains a sequence of target images; classifies each target image based on its feature information to obtain at least one set of target images of different categories, with each set containing target images of one category; performs clustering processing on the at least one set of target images of different categories to obtain feature clustering data; and optimizes the selection of target images by performing clustering processing on at least one set of target images of the same category, thereby obtaining images that represent multiple categories of the target and improving the diversity of target image selection. In addition, selecting the preferred target image based on the feature clustering data corresponding to the target can improve the accuracy of target image selection. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in this application, the accompanying drawings required in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Among them:

[0010] Figure 1 This is a flowchart illustrating the first embodiment of the image selection method of this application;

[0011] Figure 2 This application Figure 1 A flowchart illustrating an embodiment of step S12;

[0012] Figure 3 This is a schematic diagram of the storage structure of an embodiment of the classification image set that is the target of this application;

[0013] Figure 4 This application Figure 1 A flowchart illustrating an embodiment of step S13;

[0014] Figure 5 This is a schematic diagram illustrating an embodiment of the feature clustering data for the objective of this application;

[0015] Figure 6 This application Figure 1 A flowchart illustrating an embodiment of step S14;

[0016] Figure 7 This is a schematic diagram illustrating another embodiment of the feature clustering data for the objective of this application;

[0017] Figure 8 This is a flowchart illustrating the second embodiment of the image selection method of this application;

[0018] Figure 9 This is a schematic diagram of the structure of the first embodiment of the image selection device of this application;

[0019] Figure 10 This is a schematic diagram of the structure of the second embodiment of the image selection device of this application;

[0020] Figure 11 This is a schematic diagram of the structure of an embodiment of the computer device of this application;

[0021] Figure 12 This is a schematic diagram of the structure of an embodiment of the storage device of this application. Detailed Implementation

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0023] The terms "first" and "second" in this application are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such processes, methods, products, or apparatus.

[0024] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0025] Through long-term research, the inventors of this application have discovered that, taking video surveillance application scenarios as an example, multiple capture images can be obtained by capturing targets (such as pedestrians and vehicles). Since the capture process is easily affected by factors such as the target's clothing, obstruction, and environment, there may be cases where the quality of the captured images is low. Usually, one or more capture images with high image quality are selected from multiple capture images for storage based on the image quality of the captured images.

[0026] The image selection method described above results in overly homogeneous captured images of the target. In practical applications of captured images, this affects subsequent structured analysis and retrieval of the target, reducing the success rate of target retrieval.

[0027] To address the aforementioned technical problems, this application provides the following embodiments, which are described in detail below.

[0028] Please see Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the image selection method of this application. The method may include the following steps:

[0029] S11: Obtain the target image sequence.

[0030] The target can be a person, animal, object, etc. In this embodiment of the application, the target is described as a person, but the application is not limited thereto.

[0031] It should be noted that the target is the object of our interest. If we need to track and select the target image, in practical applications, we need to determine the specific physical object that the target refers to based on actual needs. For example, in scenarios such as banks and shopping malls, the target can refer to the head, face, head and shoulders, or body parts of a pedestrian; in scenarios such as roads and bridges, the target can refer to vehicles or pedestrians; in scenarios such as forests and grasslands, the target can refer to animals. This application does not impose any restrictions on this.

[0032] The target image sequence can be multiple target images containing the target. For example, it can be a target video obtained by shooting the target, and multiple video frames (target images) in the target video can be used as the target image sequence; or it can be multiple target images obtained by taking pictures of the target, and then forming the target image sequence, and so on.

[0033] In some implementations, in video surveillance scenarios, entity targets in the target video can be tracked by detecting targets (pedestrians, vehicles, motor vehicles, etc.) and the coordinate detection box of each target from multiple video frames, i.e., multiple target images. Taking pedestrians as an example, each detected target can be used as a tracking target and assigned a different ID (Identity document) value, which can be counted starting from 0 or 1.

[0034] In some embodiments, before acquiring the target image sequence of the target, at least one initial image sequence of the target may be obtained. The initial image sequence may be obtained by taking pictures of at least one target in a preset area. The preset area may be a street, shopping mall, office area, etc., and this application does not limit it.

[0035] A preset evaluation value is obtained for each target in each target image in the initial image sequence. The initial image sequence includes at least one target, and the preset evaluation value includes at least one of image quality evaluation value, target completeness evaluation value, and pose quality evaluation value. The image quality evaluation value can represent the evaluation value of the target image's exposure, sharpness, color, texture, noise, pixels, etc.; the target completeness evaluation value can represent the degree of completeness of the target in the target image; and the pose quality evaluation value can represent the evaluation value of the target's pose completeness in the target image, pose generation, etc. It is understood that the preset evaluation value may also include other evaluation values, and the preset evaluation value in this application is not limited to these.

[0036] Based on preset evaluation values, a target image sequence that meets preset preference requirements is selected from the initial image sequence. A corresponding weight value can be set for each preset evaluation value. Based on each weight value, a preset evaluation value for each target in each target image in the initial image sequence can be obtained. The preset preference requirement can be that the preset evaluation value is greater than or equal to a preset evaluation threshold.

[0037] S12: Based on the feature information of each target image in the target image sequence, classify each target image to obtain at least one classification image set of the target, and each classification image set contains target images of one classification.

[0038] The feature information of each target image in the target image sequence is extracted. In some embodiments, the feature information may include at least one of head features, head and shoulder features, decoration features, and posture features.

[0039] Based on the feature information of each target image in the target image sequence, each target image in the target image sequence is classified. The types of classification can be set according to specific application scenarios, such as the target's pose features, head and shoulder features, etc. This application does not limit this.

[0040] After classifying the target image sequence, at least one set of classified images of the target can be obtained, that is, the set of classified images of the target corresponding to each category. Each set of classified images contains target images of one category.

[0041] S13: Perform clustering processing on at least one set of classified images of the target to obtain feature clustering data.

[0042] For a target, there may be a set of images classified into at least one category. Each set of images classified is used as the target for clustering. By performing clustering on each set of images classified for the target, the feature clustering data corresponding to each set of images classified can be obtained.

[0043] In some implementations, the feature clustering data includes multiple clusters of the classified image set and related information about the clusters.

[0044] S14: Based on the feature clustering data corresponding to the target, select the preferred target image corresponding to the target.

[0045] Taking the application scenario of surveillance video as an example, the preferred target image can be selected for the target. The preferred target image refers to obtaining the preferred target image of the target in each video frame from the appearance to the disappearance of the target in the surveillance video. The preferred target image can represent the image with the highest quality expression ability in the entire life cycle of the target image sequence.

[0046] Based on the feature clustering data corresponding to the target, at least one target image can be selected from at least one image cluster in the feature clustering data of each classification image set of the target as the preferred target image corresponding to the target.

[0047] In some implementations, the target image can be used for subsequent target feature extraction, structured analysis, and retrieval, which can effectively improve the accuracy of target attribute recognition and the success rate of retrieval.

[0048] In some implementations, if the target image sequence includes multiple targets, for each target included in the target image sequence, the feature information of each target image in the target image sequence is used to classify each target image to obtain at least one classification image set for each target and subsequent steps to obtain a preferred target image for each target.

[0049] In this embodiment, a target image sequence of the target is obtained; based on the feature information of each target image in the target image sequence, each target image is classified to obtain at least one set of classified images of the target, and each set of classified images contains target images of one category; clustering is performed on the at least one set of classified images of the target to obtain feature clustering data; by performing clustering on at least one set of classified images of the same target for optimization, images representing multiple categories of the target can be obtained, improving the diversity of target image selection; in addition, based on the feature clustering data corresponding to the target, the preferred target image corresponding to the target can be selected, which can improve the accuracy of target image optimization.

[0050] In some embodiments, please refer to Figure 2 This embodiment can further extend step S12 of the above embodiment. Based on the feature information of each target image in the target image sequence, each target image is classified to obtain at least one set of classified images of the target. This embodiment may include the following steps:

[0051] S121: Extract features from each target image in the target image sequence to obtain the feature information of each target image.

[0052] This step can be referred to the specific implementation process of step S12 in the above embodiments, and will not be repeated here.

[0053] S122: Based on the feature information of the target image, classify each target image according to a preset classification strategy to obtain at least one set of classified images of the target.

[0054] The feature information may include at least one of head features, head and shoulder features, ornamentation features, and posture features. Based on the feature information of the target image, each target image is classified according to preset features, such as classifying the same target according to different posture features, to obtain a set of classified images corresponding to each preset feature of the target.

[0055] In some implementations, the preset features include the target's posture features from multiple different viewpoints. These multiple different viewpoints can be viewpoints spaced at preset angles. For example, the preset feature classification includes the target's posture features at 0 degrees, 45 degrees, 90 degrees, 135 degrees, 180 degrees, ..., 360 degrees, etc. In addition, posture features from different viewpoints can be merged into one category, such as 0-degree posture features and 180-degree posture features can be merged into one category. The classification can be carried out according to the specific application scenario. This application does not limit the preset classification strategy.

[0056] In some implementations, the classification of preset features may include at least one of frontal pose features, back pose features, and side pose features. Taking the human body as an example, frontal pose features may be features of the front of the human body, back pose features may be features of the back of the human body, and side pose features may be features of the left and / or right sides of the human body.

[0057] In some embodiments, the face containing the target is considered the frontal view. For example, the pose range (or angle range) of the frontal pose features can be 30 degrees to 150 degrees, the pose range (or angle range) of the side pose features can be 150 degrees to 210 degrees and 330 degrees (or -30 degrees) to 30 degrees, and the pose range (or angle range) of the back pose features can be 210 degrees to 330 degrees. For example, the pose range of the frontal pose features may overlap with the pose range of the side pose features, and the pose range of the back pose features may overlap with the pose range of the side pose features, etc. This application does not impose any limitations on this.

[0058] In some implementations, taking the classification of preset features as an example, including frontal pose features, back pose features, and side pose features, the target images in the target image sequence can be classified using a classification model with a preset classification strategy. The classification model can be trained using a sample image sequence, where each sample image in the sample image sequence is labeled with a corresponding classification label, such as frontal pose features, back pose features, and side pose features. The trained classification model can classify the target image sequence.

[0059] Please see Figure 3 Taking the preset features including frontal pose features, back pose features, and side pose features as an example, the target image sequence is classified according to the preset classification strategy to obtain at least one classification image set corresponding to each target. When storing at least one classification image set of the target, each target is assigned a different ID value, which is also an identifier, and is stored according to the classification image set contained in the target.

[0060] For example, target 1 includes frontal 1 (frontal pose features, the same below) and frontal 2, etc.; side 1 (side pose features, the same below) and side 2, etc.; back 1 (side pose features, the same below) and back 2, etc. Target 2 includes frontal 3 and frontal 4, etc.; side 3 and side 4, etc.; back 3 and back 4, etc. Target n includes frontal n and frontal n+1, etc.; side n and side n+1, etc.; back n and back n+1, etc., where n is a positive integer greater than 1. The classification image set of target 1 belonging to the frontal pose feature category can include frontal 1 and frontal 2, and the classification image sets of other categories are similar. In some application scenarios, target images in each classification image set of each target can also be numbered separately to manage each target image in the classification image set.

[0061] In some embodiments, please refer to Figure 4 This embodiment can further extend step S13 of the above embodiment. Clustering processing is performed on at least one set of classified images of the target to obtain feature clustering data. This embodiment may include the following steps:

[0062] S131: Use the image sets of each category of the target as cluster target data respectively.

[0063] In some implementations, at least one set of classified images contained in each target can be clustered, and each set of classified images can be used as clustering target data.

[0064] In some implementations, clustering can be performed on a target-by-target basis, using all classified image sets of the target as the clustering target data.

[0065] S132: Perform clustering processing on each cluster of target data to obtain the feature clustering data corresponding to each category.

[0066] Clustering is performed on each cluster of target data separately. That is, clustering is performed on each category of image set with the target as the unit, to obtain the feature clustering data corresponding to each category, that is, the feature clustering data corresponding to each category of image set.

[0067] The clustering method used in the clustering process can be a partition-based method, a density-based method, a hierarchical method, such as the K-means clustering algorithm and the DBSCAN clustering algorithm. This application does not impose any restrictions on this method.

[0068] Clustering is performed on each category of the target image set to obtain the feature clustering data corresponding to each category. The feature clustering data corresponding to each category includes at least one image cluster obtained by clustering the category image set corresponding to the category. That is, at least one image cluster can be obtained for each category, and each image cluster contains at least one target image.

[0069] Please see Figure 5 Taking the preset features, including classifications such as frontal pose features, side pose features, and rear pose features, as an example, at least one image cluster can be obtained for each of the frontal, side, and rear categorized image sets of the target. For example, the frontal categorized image set of target 100 may include image clusters 101, 102, and 103; the side categorized image set may include image clusters 104 and 105; and the rear categorized image set may include image cluster 106. Figure 5 In this context, each sample point in an image cluster can represent a target image.

[0070] In some embodiments, please refer to Figure 6 This embodiment can further extend step S14 of the above embodiment. Based on the feature clustering data corresponding to the target, a preferred target image corresponding to the target is selected. This embodiment may include the following steps:

[0071] S141: For each category of the target, select the cluster representation image corresponding to the category using at least one image cluster of the category.

[0072] The feature clustering data corresponding to the target includes the feature clustering data corresponding to each category of the target, and the feature clustering data corresponding to the category includes at least one image cluster obtained by clustering the category image set corresponding to the category.

[0073] For each category of each target, at least one image cluster can be obtained after the clustering of the classification image set of each category. Cluster representation images can be selected from at least one image cluster to obtain the cluster representation image corresponding to each category.

[0074] In some implementations, the number of cluster representation images corresponding to each category can be a preset number, which is a positive integer greater than or equal to 1. One or more image clusters can be selected from at least one image cluster, and then a preset number of cluster representation images can be selected from the image clusters.

[0075] In some implementations, the number of cluster representation images selected for each category can be the same as the number of image clusters, and can be cluster representation images selected from each image cluster corresponding to the category.

[0076] In some implementations, the cluster representation image for classification is the target image located at the class center of each image cluster. The class center is a special sample point in the cluster analysis used to represent a certain class, that is, it can represent an image cluster. Other sample points determine whether they belong to that image cluster by calculating their distance from it.

[0077] Please see Figure 7 For example, the classification image set for the front of target 100 may include image clusters 101, 102, and 103; the classification image set for the side may include image clusters 104 and 105; and the classification image set for the back may include image cluster 106. A clustering representation image 1001 can be selected from each image cluster. This clustering representation image 1001 represents the image cluster and represents all sample points (target images) within that image cluster.

[0078] S142: Use the clustered representation images of different target categories as the preferred target images corresponding to the target.

[0079] All cluster representation images of different categories of the target are used as the preferred target images corresponding to the target. In other words, the target image located at the center of the image cluster of each category can be used as the preferred target image corresponding to the target.

[0080] In the above scheme, for each category of the target, by utilizing at least one image cluster of the category, the cluster representation image corresponding to the category is selected, and the cluster representation images of different categories of the target are used as the preferred target images corresponding to the target. This allows the preferred target images corresponding to each category to be selected. In addition, since the cluster representation image of the category is the target image located at the center of each image cluster of the category, the selected preferred target images are representative of each category of the target, and as many preferred target images of all categories as possible can be selected.

[0081] Please see Figure 8 , Figure 8 This is a flowchart illustrating a second embodiment of the image selection method of this application. The method may include the following steps:

[0082] S21: Using the pre-stored target images of the target in the preset deduplication database, perform deduplication processing on the preferred target image corresponding to the target.

[0083] In some implementations, after step S14 of the above embodiment, that is, after selecting the preferred target image corresponding to the target based on the feature clustering data corresponding to the target, steps S21, S22 and / or S23 of this embodiment can be executed.

[0084] In some embodiments, the pre-stored target images of the target in the pre-deduplication database are stored according to categories. For example, the pre-stored target images corresponding to each category of the target are stored. This storage method can be referred to the specific implementation process of step S13 above, and will not be elaborated here.

[0085] Specifically, the preferred target image belonging to the same category can be compared with the pre-stored target image to determine whether it meets the preset deduplication requirements. For example, the similarity between the preferred target image and the pre-stored target image can be determined. If the similarity is greater than a preset similarity threshold, it can be determined that the preset deduplication requirements are met.

[0086] If the preset deduplication requirements are not met, the preferred target image will be added to the corresponding category in the preset deduplication database for storage as a subsequent pre-stored target image.

[0087] In some implementations, if the preset deduplication requirement is not met, the preferred target image of the target is added to the corresponding category of the preset deduplication database and / or the preset preferred database for storage. The preferred target images of the target in the preset preferred database are stored according to the category. The preferred target images are stored in the preset preferred database so that the preferred target images of the target can be analyzed and retrieved in the future.

[0088] In some implementations, if the preset deduplication requirement is met, the preferred target image is not stored in the preset deduplication database. Alternatively, a preset evaluation value is obtained between the preferred target image and the pre-stored target image. If the preset evaluation value of the preferred target image is higher than the preset evaluation value of the pre-stored target image, the preferred target image is stored in the corresponding category of the preset deduplication database, and the pre-stored target image is deleted. Otherwise, the preferred target image of the target is not stored in the preset deduplication database.

[0089] S22: Store the deduplicated preferred target image into a preset preferred database.

[0090] The preferred target images after deduplication can be stored in the corresponding category of the target in the preset preferred database. The preferred target images of the target in the preset preferred database are stored according to the category.

[0091] In some implementations, preferred target images that do not meet the preset deduplication requirements can be stored in the corresponding category of the target in a preset preferred database.

[0092] In some implementations, at least one preferred target image can be selected from each category of the target and stored in the corresponding category of the target in a preset preferred database.

[0093] In some implementations, the aforementioned preset deduplication database and preset preferred database can be the same database or separate databases; this application does not impose any restrictions on this.

[0094] S23: Search for the preferred target image corresponding to the target in the preset preferred database according to the preset query method; wherein, the preset query method includes at least one of the following: the image to be searched in the image search method, the identifier of the target, the attributes of the preferred target image, and the classification of the preferred target image.

[0095] After storing the preferred target image of the target in the preset preferred data, the preferred target image can be analyzed and processed, such as through retrieval and structured analysis.

[0096] The system can search for preferred target images corresponding to a target using a preset query method, i.e., perform target retrieval in a preset preferred database. The preset query method includes at least one of the following: image to be searched using image search, target identifier, attributes of preferred target images, and classification of preferred target images. It is understood that other retrieval methods can also be used, and the preset query method of this application is not limited to these.

[0097] The image search method is as follows: input the image to be searched into a preset preferred database, extract the features of the image to be searched, compare the similarity with the preferred target images in the preset preferred database, and output the N preferred target images with the highest similarity ranking, where N is a positive integer greater than or equal to 1.

[0098] The identifier of the target is: input the identifier of the target to be retrieved into the preset preferred database, compare the identifiers of the targets in the preset preferred database, and output the target that matches the identifier of the target to be retrieved and the related preferred target image.

[0099] Attributes of preferred target images: When storing preferred target images in the preset preferred database, attribute labels for preferred target images can also be set. By inputting the attributes of the preferred target images to be retrieved and comparing the attributes with the attribute labels, preferred target images that match can be output.

[0100] Classification of preferred target images: Input the classification of the preferred target images to be retrieved, compare it with the classification of preferred target images of the target stored in the preset preferred database, and output the preferred target images that match.

[0101] In some implementations, the above-mentioned multiple retrieval methods can be combined to find the preferred target image more accurately. This application does not limit the above retrieval methods.

[0102] The above solution minimizes the storage volume of preferred target images by storing the deduplicated preferred target images in a preset preferred database. In addition, when retrieving preferred target images, the search can be performed by the target's identifier. Based on the target's identifier, preferred target images of all categories can be found, which can improve the success rate and hit rate of target retrieval.

[0103] The specific implementation of this embodiment can be referred to the implementation process of the above embodiments, and will not be repeated here.

[0104] In conjunction with the above embodiments, this application also provides an image selection device. Please refer to [link / reference]. Figure 9 , Figure 9 This is a schematic diagram of the structure of the first embodiment of the image selection device of this application. The image selection device 30 includes an acquisition module 31, a classification module 32, a clustering module 33, and a selection module 34.

[0105] The acquisition module 31 is used to acquire the target image sequence of the target.

[0106] The classification module 32 is used to classify each target image according to the feature information of each target image in the target image sequence, so as to obtain at least one classification image set of the target, and each classification image set contains target images of one classification.

[0107] Clustering module 33 is used to perform clustering processing on at least one set of classified images of the target to obtain feature clustering data.

[0108] Selection module 34 is used to select the preferred target image based on the feature clustering data corresponding to the target.

[0109] The specific implementation of this embodiment can be referred to the implementation process of the above embodiments, and will not be repeated here.

[0110] In some embodiments. See Figure 9 , Figure 9 This is a schematic diagram of the structure of the second embodiment of the image selection device of this application. The image selection device 30 includes an acquisition module 31, a classification module 32, a clustering module 33, and a selection module 34. In addition, it also includes a deduplication module 35, a storage module 36, and a query module 37.

[0111] The deduplication module 35 is used to perform deduplication processing on the preferred target image corresponding to the target by using the pre-stored target image of the target in the preset deduplication database.

[0112] The storage module 36 is used to store the preferred target image after deduplication into a preset preferred database.

[0113] The query module 37 is used to search for the preferred target image corresponding to the target in the preset preferred database according to the preset query method; wherein, the preset query method includes at least one of the following: the image to be searched in the image search method, the identifier of the target, the attributes of the preferred target image, and the classification of the preferred target image.

[0114] The specific implementation of this embodiment can be referred to the implementation process of the above embodiments, and will not be repeated here.

[0115] Regarding the above embodiments, this application provides a computer device; please refer to [link / reference]. Figure 11 , Figure 11 This is a schematic diagram of the structure of a computer device according to an embodiment of the present application. The computer device 40 includes a memory 41 and a processor 42, wherein the memory 41 and the processor 42 are coupled to each other. The memory 41 stores program data, and the processor 42 is used to execute the program data to implement the steps of any embodiment of the image selection method described above.

[0116] In this embodiment, processor 42 can also be referred to as a CPU (Central Processing Unit). Processor 42 may be an integrated circuit chip with signal processing capabilities. Processor 42 can also be a general-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component. The general-purpose processor can be a microprocessor, or processor 42 can be any conventional processor.

[0117] The specific implementation of this embodiment can be referred to the implementation process of the above embodiments, and will not be repeated here.

[0118] The methods described in the above embodiments can be implemented as computer programs; therefore, this application proposes a storage device. Please refer to [link to relevant documentation]. Figure 12 , Figure 12This is a schematic diagram of a storage device according to an embodiment of the present application. The storage device 50 stores program data 51 that can be executed by a processor. The program data 51 can be executed by the processor to implement the steps of any embodiment of the image selection method described above.

[0119] The specific implementation of this embodiment can be referred to the implementation process of the above embodiments, and will not be repeated here.

[0120] In this embodiment, the storage device 50 can be a medium that can store program data 51, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk. Alternatively, it can be a server that stores the program data 51. The server can send the stored program data 51 to other devices for execution, or it can run the stored program data 51 itself.

[0121] In the several embodiments provided in this application, it should be understood that the disclosed methods and apparatus can be implemented in other ways. For example, the apparatus implementations described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0122] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0123] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0124] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage device, which is a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application.

[0125] Obviously, those skilled in the art should understand that the modules or steps of this application described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device, or fabricating them separately as individual integrated circuit modules, or fabricating multiple modules or steps as a single integrated circuit module. Thus, this application is not limited to any particular combination of hardware and software.

[0126] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. An image selection method, characterized in that, The method includes: Obtain a preset evaluation value for each target in each target image in the initial image sequence; wherein, the preset evaluation value includes a pose quality evaluation value; Based on the preset evaluation value, a target image sequence that meets the preset preference requirements is selected from the initial image sequence; Based on the feature information of each target image in the target image sequence, each target image is classified to obtain at least one set of classified images of the target. Each set of classified images contains one type of target image, including: Feature extraction is performed on each target image in the target image sequence to obtain the feature information of each target image; Based on the feature information of the target image, each target image is classified according to preset features to obtain the classification image set corresponding to each preset feature of the target; wherein, the preset features include the pose features of the target under multiple different viewpoints; Clustering processing is performed on at least one categorized image set of the target to obtain feature clustering data; wherein, the feature clustering data corresponding to the target includes feature clustering data corresponding to each category of the target, and the feature clustering data corresponding to each category includes at least one image cluster obtained by clustering the categorized image set corresponding to the category; Based on the feature clustering data corresponding to the target, a preferred target image corresponding to the target is selected, including: For each category of the target, a clustering representation image corresponding to the category is selected using at least one image cluster of the category; wherein, the clustering representation image of the category is the target image located at the center of each image cluster of the category; The cluster representation images of different classifications of the target are used as the preferred target images corresponding to the target.

2. The method according to claim 1, characterized in that, The preset features include at least one of the following: frontal posture features, back posture features, and side posture features.

3. The method according to claim 1, characterized in that, The clustering process performed on at least one classified image set of the target to obtain feature clustering data includes: Each of the classified image sets of the target is used as the clustering target data; Clustering processing is performed on each of the cluster target data of the target to obtain the feature clustering data corresponding to each category.

4. The method according to claim 1, characterized in that, After selecting the preferred target image based on the feature clustering data corresponding to the target, the method further includes: Using the pre-stored target images of the target in the preset deduplication database, the preferred target images corresponding to the target are deduplicated; The preferred target image after deduplication is stored in a preset preferred database.

5. The method according to claim 4, characterized in that, The pre-stored target images of the target in the preset deduplication database are stored according to the classification. The step of using pre-stored target images of the target in a preset deduplication database to perform deduplication processing on the preferred target image corresponding to the target includes: The preferred target image of the target belonging to the same category is compared with the pre-stored target image of the target to determine whether the preset deduplication requirement is met. If the preset deduplication requirements are not met, the preferred target image of the target will be added to the corresponding category of the preset deduplication database and / or the preset preferred database.

6. The method according to claim 4, characterized in that, After storing the deduplicated preferred target image into a preset preferred database, the method further includes: Search the preset preferred database for the preferred target image corresponding to the target according to a preset query method; The preset query method includes at least one of the following: the image to be searched using image search, the identifier of the target, the attributes of the preferred target image, and the classification of the preferred target image.

7. The method according to claim 1, characterized in that, The initial image sequence includes at least one target, and the preset evaluation value also includes at least one of image quality evaluation value and target integrity.

8. A computer device, characterized in that, The method includes a memory and a processor coupled to each other, the memory storing program data and the processor executing the program data to implement the steps of the method according to any one of claims 1 to 7.

9. A storage device, characterized in that, The system stores program data that can be executed by a processor, the program data being used to implement the steps of the method according to any one of claims 1 to 7.

Images