Target clustering method, electronic equipment and storage medium

Abstract

The invention discloses a target clustering method, electronic equipment and a storage medium, and the method comprises the steps: obtaining the spatial information and time information of image collection points corresponding to at least a part of to-be-processed images, and determining a preset partition of each image collection point through the spatial information and time information; performing target clustering on the at least part of the to-be-processed image to obtain clustering information corresponding to each target in the at least part of the to-be-processed image; in response to the fact that the clustering information of the at least one target is adjusted, obtaining a corresponding adjustment record table, and adjusting a preset partition by using the adjustment record table to determine an expected partition of each image acquisition point; performing partition classification on all the to-be-processed images according to the expected partitions; according to the method, the preset partition is determined by using the time information and the space information of the image acquisition points, and then the target cluster is subjected to comparison before and after adjustment, so that the preset partition is adjusted, the expected partition is obtained, the target partition of the to-be-processed image is effectively determined, and the clustering accuracy is improved.

Description

Technical Field

[0001] The present application relates to the technical field of image clustering, and in particular to a target clustering method, electronic device, and storage medium. Background Art

[0002] With the explosive growth of computer applications and the continuous advancement of target clustering technology, person-based clustering based on face recognition technology has gradually become important information in certain social fields.

[0003] Current target clustering techniques use deep learning to extract feature vectors from small image segments of the target. These vectors are then used to calculate the similarity between these vectors. However, due to factors such as the distance, angle, and lighting conditions at the time the image acquisition device captures the target, some images of the same target may not meet the clustering criteria. This can lead to multiple images of the same target, compromising clustering accuracy. Therefore, effectively improving the recall rate of target clustering is an urgent issue. Summary of the Invention

[0004] The technical solution to the technical problem mainly solved by this application is to provide a target clustering method, electronic device and storage medium, which are applicable to more scenarios, reduce the occurrence of false detection and omission, and can effectively improve clustering accuracy.

[0005] In order to solve the above technical problems, a technical solution adopted in the present application is: to provide a target clustering method, including: obtaining spatial information and time information of image acquisition points corresponding to at least part of the image to be processed, and using the spatial information and the time information to determine the preset partition of each of the image acquisition points; performing target clustering on at least part of the image to be processed to obtain clustering information corresponding to each target in at least part of the image to be processed; in response to the clustering information of at least one of the targets being adjusted, obtaining a corresponding adjustment record table, and using the adjustment record table to adjust the preset partition to determine the expected partition of each of the image acquisition points; and partitioning and classifying all the images to be processed according to the expected partition.

[0006] In some embodiments, the acquiring of spatial information and temporal information of image acquisition points corresponding to at least a portion of the image to be processed, and determining a preset partition of each of the image acquisition points using the spatial information and the temporal information, includes: acquiring multiple image acquisition points corresponding to at least a portion of the image to be processed, and acquiring spatial information and temporal information of each of the image acquisition points; and grouping the multiple image acquisition points using the spatial information and the temporal information to determine a preset partition of each of the image acquisition points, wherein each of the preset partitions includes at least one of the image acquisition points.

[0007] In some embodiments, it also includes: obtaining the original image information and the image acquisition point information of each of the at least part of the images to be processed as the point information of each of the images to be processed; in response to the consistency between the point information and the spatial information, using the preset partition corresponding to the consistent information as the current image preset partition of the image to be processed.

[0008] In some embodiments, the target clustering of the at least portion of the image to be processed to obtain clustering information corresponding to each target in the at least portion of the image to be processed includes: obtaining at least one target in the at least portion of the image to be processed; and performing target clustering on the image to be processed containing the target to obtain clustering information corresponding to each target.

[0009] In some embodiments, in response to the clustering information of at least one of the targets being adjusted, a corresponding adjustment record table is obtained, and the preset partition is adjusted using the adjustment record table to determine the expected partition of each of the image acquisition points, including: in response to the clustering information of at least one of the targets being adjusted, the adjustment record table is determined with the adjustment information of the clustering information; and the preset partition of each of the image acquisition points is adjusted using the adjustment information in the adjustment record table to determine the expected partition of each of the image acquisition points.

[0010] In some embodiments, in response to the clustering information of at least one of the targets being adjusted, the adjustment record table is determined based on the adjustment information of the clustering information, including: obtaining the clustering information of each of the targets, and determining the adjustment information corresponding to each of the clustering information; using the adjustment information to obtain the image that needs to be adjusted corresponding to each of the targets; adjusting the image that needs to be adjusted to the correct clustering information of the target, and obtaining the adjustment record table.

[0011] In some embodiments, adjusting the preset partition of each image acquisition point using the adjustment information in the adjustment record table to determine the desired partition of each image acquisition point includes: traversing the adjustment record table of the image to be processed to obtain image acquisition point information, partition information before adjustment, and partition information after adjustment for each recorded image; merging the partition information before adjustment and the partition information after adjustment corresponding to the same image point information into an array to determine an image acquisition point adjustment list; traversing the image acquisition point adjustment list to obtain the number of records of the traversed current image acquisition point in the image adjustment record table, and obtaining the number of images in the image to be processed whose point information is the current image acquisition point; in response to a ratio of the number of records to the number of images being greater than a preset ratio, obtaining the partition information after adjustment of the current image acquisition point in the array from the image acquisition point adjustment list, and determining a partition information table after adjustment; traversing the partition information table after adjustment to obtain a partition list for the current image acquisition point, and determining a partition containing more than a preset number of images to be processed as the desired partition for the current image acquisition point.

[0012] In some embodiments, the record corresponding to the current image acquisition point is obtained in the preset partition, and the preset partition in the record is updated to the expected target partition of the current image acquisition point, wherein the current image acquisition point is the image acquisition point of the expected target partition traversed; the original image information and the image acquisition point information of each of the images to be processed are obtained, and an image information list is determined; the image information list is traversed to determine the partition information of the updated expected partition of each of the images to be processed, and use it as the corresponding image target partition of the image to be processed.

[0013] To solve the above technical problems, another technical solution adopted in this application is: to provide an electronic device, which includes a memory and a processor coupled to the memory, and the memory stores at least one computer program, and when the at least one computer program is loaded and executed by the processor, it is used to implement the target clustering method as mentioned above.

[0014] To solve the above technical problems, another technical solution adopted in this application is: providing a computer-readable storage medium, wherein the computer-readable storage medium has at least one program, and when the at least one program is loaded and executed by the processor, it is used to implement the target clustering method as described above.

[0015] Different from the current technology, the target clustering method provided by the present application includes: obtaining spatial information and time information of image acquisition points corresponding to at least part of the image to be processed, and using the spatial information and time information to determine the preset partition of each image acquisition point; performing target clustering on at least part of the image to be processed to obtain clustering information corresponding to each target in at least part of the image to be processed; in response to the clustering information of at least one target being adjusted, obtaining the corresponding adjustment record table, and using the adjustment record table to adjust the preset partition to determine the expected partition of each image acquisition point; partitioning and classifying all images to be processed according to the expected partition; that is, in the present application, the preset partition is first determined using the time information and spatial information of the image acquisition points of a small number of images to be processed, and then the clustering information before and after the target clustering adjustment is analyzed and compared to obtain the adjustment record table, and then the preset partition is adjusted using the adjustment record table to obtain the expected partition of each image acquisition point, effectively determining the optimal target partition of the image to be processed, and improving the clustering accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without inventive efforts. Among them:

[0017] Figure 1 This is a flow chart of an embodiment of a target clustering method in this application;

[0018] Figure 2 This is a schematic structural diagram of an embodiment of a target clustering system in the present application;

[0019] Figure 3 This is a schematic structural diagram of an embodiment of an electronic device in this application;

[0020] Figure 4 It is a structural diagram of an embodiment of a computer-readable storage medium in this application. DETAILED DESCRIPTION

[0021] The present invention will be described in further detail below with reference to the accompanying drawings and examples. It is particularly noted that the following examples are intended only to illustrate the present invention and are not intended to limit the scope of the present invention. Similarly, the following examples are only some embodiments of the present invention and are not intended to be all embodiments. All other embodiments obtained by those of ordinary skill in the art without creative effort are intended to fall within the scope of protection of the present invention.

[0022] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute a separate or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0023] Traditional target clustering techniques use deep learning to extract feature vectors from small target images within an image, then calculate the similarity between these vectors to cluster the targets. However, due to factors such as distance, angle, and lighting when the image acquisition device captures the target, the similarity of some images of the same target may not meet the clustering criteria, resulting in multiple images of the same target. Therefore, effectively improving the recall rate of target clustering is an urgent problem to be solved.

[0024] Therefore, a target clustering method is provided, which utilizes the spatiotemporal information of a small number of image acquisition points of the image to be processed to determine the preset partitions, perform target clustering, determine the clustering information corresponding to each target in the image to be processed, and for the clustering information that needs to be adjusted, determine the adjustment record table, and then adjust the preset partitions to obtain the expected partitions for each image acquisition point, and then partition and classify all the images to be processed according to the expected partitions, effectively determine the optimal target partitions of the images to be processed, and improve the clustering accuracy.

[0025] See also Figure 1 , Figure 1 This is a flow chart of an embodiment of the target clustering method in this application; it should be noted that if there are substantial results, the method of this application is not based on Figure 1 The process sequence shown is limited.

[0026] like Figure 1 As shown, the target clustering method of the present application may include the following steps.

[0027] S10. Obtain spatial information and temporal information of image acquisition points corresponding to at least part of the image to be processed, and determine a preset partition of each image acquisition point using the spatial information and the temporal information.

[0028] Among them, the image to be processed refers to the image that has not yet been clustered, which can be a historical image or an image acquired in real time; the image acquisition point refers to the location of the image acquisition device; spatial information refers to various types of data and information related to the geographic space location, mainly describing the position, shape, distribution, attributes and dynamic changes of objects on the earth's surface or near-earth space, such as the position, shape, distribution, attributes and dynamic changes of the image acquisition device; time information refers to the specific time when the image acquisition device acquires the image to be processed; preset partition refers to the division of the image acquisition point in combination with spatial information and time information to obtain the initial partition of the image acquisition point, that is, the preset partition.

[0029] Specifically, a plurality of images to be processed are acquired through an image acquisition device, and an image acquisition point of each image to be processed is acquired, and then the images to be processed are randomly sampled to determine sample image data, that is, at least part of the images to be processed, and spatial information and time information of the image acquisition points corresponding to at least part of the images to be processed are acquired; and then, the image acquisition points corresponding to at least part of the images to be processed are partitioned according to the dimension of spatial information and the dimension of time information to obtain preset partitions.

[0030] In some embodiments, each preset partition includes at least one image acquisition point, that is, each preset partition includes at least one image acquisition device.

[0031] S20 , performing object clustering on at least a portion of the image to be processed to obtain clustering information corresponding to each object in at least a portion of the image to be processed.

[0032] Among them, target clustering refers to the process of grouping objects, data points or entities with similar characteristics or common goals; its core purpose is to simplify complex sets into several meaningful subsets by analyzing characteristics or behaviors. Here, the target clustering is performed on the targets in the image to be processed; clustering information refers to the subset of targets that appear in the image to be processed.

[0033] S30 . In response to the clustering information of at least one target being adjusted, obtain a corresponding adjustment record table, and use the adjustment record table to adjust the preset partition to determine a desired partition of each image acquisition point.

[0034] Among them, the clustering information is adjusted, which means that the clustering information of the target may have clustering errors or omissions, so the images with clustering errors or omissions need to be adjusted to the correct clustering information; the adjustment record table refers to the list composed of records in the adjustment process; the expected partition refers to the target partition corresponding to the image acquisition point.

[0035] Specifically, when there is adjustment in the clustering information of at least one target, the records of the adjustment process are obtained, and an adjustment record table is composed of all the records. Through the corresponding records in the adjustment record table, the preset partitions of the image acquisition points with existing records are adjusted according to the records in the adjustment record table, that is, the corresponding preset partitions are adjusted to the expected partitions after adjustment in the adjustment record table.

[0036] S40: partition and classify all the images to be processed according to the desired partitions.

[0037] Partition classification refers to clustering all images to be processed.

[0038] Specifically, after obtaining the desired partition of each image acquisition point, all the images to be processed are partitioned and classified, thereby obtaining the final clustering information corresponding to all the images to be processed.

[0039] In this embodiment, the spatiotemporal information of a small number of image acquisition points of the images to be processed is used to determine preset partitions, perform target clustering, and determine the clustering information corresponding to each target in the image to be processed. For the clustering information that needs to be adjusted, an adjustment record table is determined, and then the preset partitions are adjusted to obtain the expected partitions for each image acquisition point. All images to be processed are then partitioned and classified according to the expected partitions, effectively determining the optimal target partitions of the images to be processed and improving clustering accuracy.

[0040] In some embodiments, S10 acquires spatial information and temporal information of at least a portion of image acquisition points corresponding to the image to be processed, and determines a preset partition of each image acquisition point using the spatial information and temporal information, which may include the following operations.

[0041] First, a plurality of image acquisition points corresponding to at least a portion of the image to be processed are obtained, and spatial information and time information of each image acquisition point are obtained.

[0042] There may be multiple image acquisition points, corresponding to the position information of the image acquisition device, and the multiple image acquisition points here are also the image acquisition points corresponding to all the images to be processed.

[0043] Specifically, random sampling is performed from all the images to be processed to obtain at least part of the images to be processed, and the image acquisition points corresponding to at least part of the images to be processed are the image acquisition points corresponding to all the images to be processed; then, the spatial information and time information corresponding to each image acquisition point are extracted, that is, the spatial information corresponding to the image acquisition device lock and the time information when the image to be processed is acquired.

[0044] Then, the plurality of image acquisition points are grouped using the spatial information and the temporal information to determine a preset partition of each image acquisition point, wherein each preset partition includes at least one image acquisition point.

[0045] The preset partition refers to dividing the image acquisition points in combination with spatial information and time information to obtain initial partitions of the image acquisition points, namely, the preset partitions.

[0046] Specifically, after obtaining the spatial information and time information of each image acquisition point, the image acquisition points corresponding to at least part of the image to be processed are grouped according to the dimension of the spatial information and the dimension of the time information, that is, partitioned to obtain the preset partition corresponding to each image acquisition point.

[0047] In some embodiments, each preset partition includes at least one image acquisition point, that is, includes at least one image acquisition device. Each preset partition may include one or more image acquisition devices.

[0048] In this embodiment, the image acquisition points are partitioned according to the two dimensions of spatial information and temporal information, and the preset partitions of each image acquisition point are determined. The cluster partitioning that matches the image to be processed can be completed without the need for historical clustering information, which is conducive to improving the target clustering recall rate, that is, increasing the proportion of samples that are correctly identified and classified as the same category by the model in the clustering task.

[0049] Furthermore, the following operations may be included.

[0050] Original image information and image acquisition point information of at least a portion of the images to be processed are obtained as point position information of each image to be processed.

[0051] Among them, the original image information refers to the basic information of the image to be processed, such as the acquisition time information of the image to be processed, the target information contained in the image to be processed, the ID of the image to be processed, etc.; the image acquisition point information refers to the partition information where the image to be processed is located; the point information includes the above-mentioned acquisition time information, target information, the ID of the image to be processed and the image acquisition point information, etc.

[0052] Specifically, after randomly sampling to obtain at least part of the images to be processed, the original image information and the image acquisition point information corresponding to each image to be processed in the sample are obtained, and then the original image information and the image acquisition point information are used as the point information of each image to be processed.

[0053] In response to the point information and the spatial information being consistent, the preset partition corresponding to the consistent information is used as the image preset partition of the current image to be processed.

[0054] Among them, the point information of the image to be processed is consistent with the spatial information of the image acquisition point, indicating that the image to be processed and the image acquisition point are located in the same preset partition. Therefore, the preset partition corresponding to the consistency of the point information and the spatial information of the image acquisition point can be used as the image preset partition of the current image to be processed. The image preset partition here is the partition corresponding to each image to be processed in the sample.

[0055] In this embodiment, the point information and the spatial information are compared to determine the preset image partition of each image to be processed in the sample, providing a basis for subsequent processing.

[0056] In some embodiments, S20 performs object clustering on at least a portion of the image to be processed to obtain clustering information corresponding to each object in at least a portion of the image to be processed, which may include the following operations.

[0057] First, at least one object in at least a portion of the image to be processed is acquired.

[0058] Each image to be processed may contain at least one target, where the target refers to the target object corresponding to the retrieval, such as pedestrians, animals, plants, objects, etc. that appear in the image.

[0059] Specifically, target detection is performed on at least a portion of the image to be processed to obtain at least one target in the image to be processed.

[0060] Then, target clustering is performed on the image to be processed containing the target to obtain cluster information corresponding to each target.

[0061] Among them, target clustering refers to clustering the targets appearing in the image to be processed.

[0062] Specifically, after obtaining the target in the image to be processed, target clustering processing is performed on the image to be processed containing the target, and all images to be processed containing the current target are determined as clustering information corresponding to the current target.

[0063] In some embodiments, target clustering may be performed separately on each preset partition to obtain clustering information containing the current target in each preset partition.

[0064] For example, the preset partitions are A, B, and C. There are ten images to be processed in each preset partition. For target D, there are five images to be processed in zone A that contain target D, and these five images to be processed are used as clustering information of zone A; there are two images to be processed in zone B that contain target D, and these two images to be processed are used as clustering information of zone B; there are three images to be processed in zone C that contain target D, and these three images to be processed are used as clustering information of zone C.

[0065] In this embodiment, by performing target clustering on the images to be processed in the sample, initial clustering information can be obtained. Subsequently, the changes before and after the clustering information adjustment are analyzed and processed to determine an adjustment record table, which provides a basis for the subsequent steps.

[0066] In some embodiments, S30, in response to the clustering information of at least one target being adjusted, obtaining a corresponding adjustment record table, and adjusting the preset partition using the adjustment record table to determine the desired partition of each image acquisition point, may include the following operations.

[0067] First, in response to clustering information of at least one target being adjusted, an adjustment record table is determined using adjustment information of the clustering information.

[0068] The clustering information is adjusted, which means that after clustering the targets in the to-be-processed image in the sample, the obtained clustering information is adjusted, such as adjusting the incorrectly clustered or omitted targets to the correct clustering information.

[0069] Specifically, after obtaining the clustering information of the targets of the image to be processed in the sample, the clustering information is analyzed to determine the images to be processed corresponding to the clustering errors and omitted targets, and adjust them to the correct clustering information; in response to the clustering information of at least one target being adjusted, the adjustment information of the clustering information during the adjustment process is obtained, and the adjustment information of the clustering information is recorded in the adjustment record table.

[0070] It is understandable that the clustering information can be analyzed through an analysis module or through manual analysis to determine clustering errors and missed targets, and then the images to be processed corresponding to the clustering errors and missed targets are adjusted to the correct clustering information.

[0071] The preset partition of each image acquisition point is adjusted using the adjustment information in the adjustment record table to determine the desired partition of each image acquisition point.

[0072] The adjustment record table records the ID of the image being adjusted, the ID of the image acquisition point to which the image belongs, the ID of clustering information before adjustment, and the ID of clustering information after adjustment.

[0073] Specifically, after obtaining the adjustment record table and adjusting the to-be-processed images corresponding to clustering errors and missed targets to the correct clustering information, the preset partitions of the corresponding image acquisition points are also adjusted, thereby determining the desired partitions of each image acquisition point.

[0074] It is understandable that adjusting only the images to be processed recorded in the adjustment record table, rather than adjusting all the images to be processed, can effectively reduce the difficulty of adjustment and save costs.

[0075] Furthermore, in response to the clustering information of at least one target being adjusted, determining the adjustment record table using the adjustment information of the clustering information may further include the following operations.

[0076] The clustering information of each target is obtained, and the adjustment information corresponding to each clustering information is determined.

[0077] The adjustment information refers to the adjustment information obtained by adjusting the target clustering information when there are clustering errors or omissions.

[0078] Specifically, the clustering information of each target in the image to be processed in the sample is obtained, and the clustering errors or missed targets are determined through analysis by the analysis module or manual analysis. Then, the image to be processed corresponding to the clustering errors and missed targets is adjusted to the correct clustering information, and the adjustment information corresponding to each clustering information is determined.

[0079] The adjustment information is used to obtain the image that needs to be adjusted corresponding to each target.

[0080] The adjustment information includes the image ID of the image to be adjusted, so the image to be adjusted corresponding to each target in the image to be processed can be determined. The image to be adjusted is the image to be processed with clustering errors or missing targets.

[0081] Adjust the image that needs to be adjusted to the correct clustering information of the target and obtain an adjustment record table.

[0082] The correct clustering information may be the preset clustering information determined by the analysis module, or the preset clustering information determined by manual analysis.

[0083] Specifically, after obtaining the image that needs to be adjusted, the image that needs to be adjusted is adjusted to the correct clustering information of the target based on the preset clustering information determined by analysis, and then the information during the adjustment process is used as an adjustment record table; wherein, in the adjustment record table, each record contains: the ID of the image to be adjusted, the ID of the image acquisition point to which the image belongs, the clustering information ID before adjustment, and the clustering information ID after adjustment.

[0084] Furthermore, adjusting the preset partition of each image acquisition point using the adjustment information in the adjustment record table to determine the desired partition of each image acquisition point may also include the following operations.

[0085] First, the image adjustment record table is traversed to obtain the image acquisition point information, the partition information before adjustment, and the partition information after adjustment of each recorded image.

[0086] Among them, the image adjustment record table is the adjustment record table, the image acquisition point information refers to the image acquisition point ID to which the image belongs; the partition information before adjustment refers to the cluster information ID before adjustment, and the partition information after adjustment refers to the cluster information ID after adjustment.

[0087] Specifically, the comparison module traverses the image adjustment record table, and then obtains the image acquisition point ID, the pre-adjustment clustering information ID, and the post-adjustment clustering information ID corresponding to each record in the image adjustment record table.

[0088] Next, the subarea information before adjustment and the subarea information after adjustment corresponding to the same image acquisition point information are combined into an array to determine an image acquisition point adjustment list.

[0089] The same image acquisition point information is the same image acquisition point ID.

[0090] Specifically, the image acquisition point ID is deduplicated, that is, different clustering information IDs before adjustment and clustering information IDs after adjustment corresponding to the same image acquisition point ID are merged into an array, that is, the array members are information pairs consisting of the clustering information ID before adjustment and the clustering information ID after adjustment, and then an image acquisition point adjustment list is generated; wherein each information in the image acquisition point adjustment list includes: the image acquisition point ID, an array of one or more information pairs.

[0091] Next, the image acquisition point adjustment list is traversed to obtain the number of records of the traversed current image acquisition point in the image adjustment record table, and the number of images whose point information in the image to be processed is the current image acquisition point is obtained.

[0092] The image acquisition point traversed during the process of traversing the image acquisition point adjustment list is used as the current image acquisition point.

[0093] Specifically, after obtaining the image acquisition point adjustment list, the comparison module traverses the image acquisition point adjustment list, counts the number of records of the current image acquisition point in the image adjustment record table, and then counts the point information from the image to be processed as the image data of the current image acquisition point.

[0094] Next, in response to the ratio of the number of records to the number of images being greater than a preset ratio, the adjusted partition information of the array of the current image acquisition point is obtained from the image acquisition point adjustment list, and an adjusted partition information table is determined.

[0095] The preset ratio can be set according to actual conditions, for example, the preset ratio is set to 1 / 2.

[0096] Specifically, after obtaining the number of records and the number of images, the ratio between the number of records and the number of images is compared. In response to the ratio of the number of records to the number of images being greater than a preset ratio, the array of information pairs of the current image acquisition point is searched from the image acquisition point adjustment list, and all adjusted partition information in the array of the current image acquisition point is obtained, that is, all adjusted clustering information IDs in the array of the current image acquisition point are obtained and saved as an adjusted partition information table.

[0097] It is understandable that if the ratio of the number of recordings to the number of images is less than or equal to the preset ratio, the process returns to the previous step and processes the next image acquisition point.

[0098] Then, the adjusted partition information table is traversed to obtain a partition list of the current image acquisition point, and a partition containing images to be processed that is greater than a preset image quantity threshold is used as the desired partition of the current image acquisition point.

[0099] The preset image quantity threshold can be set according to actual conditions.

[0100] Specifically, after obtaining the adjusted partition information table, the partition list of the current image acquisition point is obtained, and then the partition containing the number of images to be processed greater than a preset image number threshold is used as the desired partition of the current image acquisition point.

[0101] In some embodiments, the following operations may also be included.

[0102] The array of information pairs of the current image acquisition point is searched from the partition information table after adjustment to obtain all the adjusted cluster information IDs in the value and save them as the adjusted cluster information list.

[0103] Traverse the adjusted cluster information list and use the traversed cluster information as the current cluster information.

[0104] Because the preset partitions output by the preset partition generation module include the image ID and the partition to which the image belongs, by traversing the images in the current clustering information, a list of partitions in the current clustering information is counted. Each item in the list contains the partition and the number of images contained in the partition.

[0105] After traversing the adjusted clustering information list of the current image acquisition point, the partition lists obtained are merged, and the items with the same partition are combined into one item, and the number of images contained is accumulated. In this way, the partition list of the current image acquisition point is obtained.

[0106] In the partition list of the current image acquisition point, the partition containing the largest number of images is found as the desired partition for the current image acquisition point. Therefore, after traversing the image acquisition point adjustment list, the desired partitions for all image acquisition points can be obtained, and a list of desired target partitions for image acquisition points can also be generated.

[0107] In this embodiment, the comparison module is used to compare the information before and after adjustment to obtain the corresponding adjustment record table, and the adjustment record table is used to adjust the preset partition to determine the expected partition of each image acquisition point, thereby improving the accuracy of the partition and providing accurate data for subsequent target clustering.

[0108] Furthermore, the following operations may be included.

[0109] Obtain the record corresponding to the current image acquisition point in the preset partition, and update the preset partition in the record to the desired partition of the current image acquisition point;

[0110] The current image acquisition point is the image acquisition point of the desired partition that has been traversed, and the current image acquisition point is the image acquisition point of the preset partition that needs to be adjusted.

[0111] Specifically, the preset partition contains a record of each image acquisition point, so the record corresponding to the current image acquisition point can be obtained in the preset partition, and then the preset partition of the current image acquisition point in the record can be updated to the expected partition of the current image acquisition point according to the adjustment record table.

[0112] The original image information and the image acquisition point information of each image to be processed are obtained, and an image information list is determined.

[0113] The original image information includes the image ID and the image acquisition point information.

[0114] Specifically, the image ID and image acquisition point information of each image to be processed are obtained from all images to be processed, and then the image ID and image acquisition point information are used as the image information corresponding to the image to be processed, thereby determining the image information list corresponding to all images to be processed.

[0115] The image information list is traversed to determine the partition information of the expected partition of each image to be processed after the update, and the partition information is used as the image target partition of the corresponding image to be processed.

[0116] The target image partition refers to the partition corresponding to the image to be processed after adjustment.

[0117] Specifically, after obtaining the image information list, the image information list is traversed, and the partition information of the expected partition of each image to be processed after the update is determined from the image information list based on the image ID and the image acquisition point information, and used as the image target partition of the corresponding image to be processed.

[0118] In some embodiments, S40 partitions and classifies all to-be-processed images according to desired partitions, which may include the following operations.

[0119] First, obtain all the objects in the image to be processed.

[0120] Each image to be processed may contain at least one target, where the target refers to the target object corresponding to the retrieval, such as pedestrians, animals, plants, objects, etc. that appear in the image.

[0121] Specifically, target detection is performed on all images to be processed to obtain at least one target in the images to be processed.

[0122] Then, the desired partition is used to perform a second target clustering on the image to be processed containing the target, so as to obtain the final clustering information corresponding to each target in the image to be processed.

[0123] Among them, target clustering refers to clustering the targets appearing in the image to be processed.

[0124] Specifically, after obtaining the target in the image to be processed, target clustering processing is performed on the image to be processed containing the target, and all images to be processed containing the current target are determined as final clustering information corresponding to the current target.

[0125] In this embodiment, the spatiotemporal information of a small number of image acquisition points of the images to be processed is used to determine preset partitions, perform target clustering, and determine the clustering information corresponding to each target in the image to be processed. For the clustering information that needs to be adjusted, an adjustment record table is determined, and then the preset partitions are adjusted to obtain the expected partitions for each image acquisition point. All images to be processed are then partitioned and classified according to the expected partitions, effectively determining the optimal target partitions of the images to be processed and improving clustering accuracy.

[0126] This application also provides a target clustering system.

[0127] See also Figure 2 , Figure 2 It is a structural diagram of an embodiment of a target clustering system in this application.

[0128] like Figure 2 As shown, the target clustering system 500 includes: a preset partition generation module 510 , a target clustering module 520 , an adjustment module 530 , a comparison module 540 , and a dynamic partition generation module 550 .

[0129] The preset partition generation module 510 is used to visualize all image acquisition devices based on spatial and temporal information. All image acquisition devices are then partitioned, both spatially and temporally, with each partition containing one or more image acquisition devices. The results of the image acquisition device partitioning operation are saved as "image acquisition device preset partitions." Next, based on the consistency between the point information carried in the sample data and the spatial information of the image acquisition device, "image preset partitions" are generated for the image to be processed of the sample data. The sample data is at least a portion of all the images to be processed.

[0130] The target clustering module 520 performs target clustering on the image to be processed of the sampled sample data based on the “image preset partition” information generated by the “preset partition generation module” to obtain clustering information corresponding to each target in at least part of the image to be processed.

[0131] Adjustment module 530 is configured to obtain a corresponding adjustment record table in response to adjustments to the clustering information of at least one target. For example, the clustering information obtained after target clustering can be adjusted, either manually or mechanically, to adjust incorrectly or omitted targets to the correct cluster information. During this adjustment process, an image adjustment record list is output. Each record contains the ID of the image being adjusted, the ID of the image acquisition point, the ID of the cluster information before adjustment, and the ID of the cluster information after adjustment. This image adjustment record list serves as calibration information for subsequent partition adjustments.

[0132] Comparison module 540 compares the information before and after the adjustment. For example, it analyzes whether the original images contained in the image acquisition points in each adjustment record are correctly or incorrectly archived by the target clustering algorithm. Based on the ratio of the original images to the original images, it determines whether the partition to which the image acquisition points belong needs to be modified. The partition containing the image acquisition points to which the original images belong, under the adjusted clustering information ID, is used as the desired partition for the image acquisition points in the record.

[0133] The dynamic partition generation module 550 uses the adjustment record table to adjust the preset partitions to determine the desired partition for each image acquisition point. For example, based on the desired partition record for the image acquisition point, the partition of the image acquisition device to be adjusted in the "preset partition of the image acquisition device" is adjusted to the desired partition. This adjusted partition information is called the "desired target partition of the image acquisition point." Based on the consistency between the point information carried in the original image and the spatial information of the image acquisition device, "image target partitions" are generated for all original images. The target clustering module then clusters all original image data based on the "image target partitions" to obtain the final clustering results.

[0134] In this application, an electronic device is also provided.

[0135] See Figure 3 , Figure 3 FIG. 1 is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. The electronic device can execute the steps of the target clustering method in the above method.

[0136] The electronic device 600 includes a memory 620, a processor 610 coupled to the memory, and at least one computer program stored in the memory 620 and executable on the processor 610. When the processor 610 loads and executes the at least one computer program, it is configured to implement the target clustering method steps described above. For details, please refer to the detailed description of the method described above and will not be repeated here.

[0137] This application also includes a computer-readable storage medium.

[0138] See also Figure 4 , Figure 4 It is a structural diagram of an embodiment of a computer-readable storage medium in this application.

[0139] The computer-readable storage medium 700 stores at least one program 710. When the at least one program 710 is loaded and executed by the processor, it is used to implement the steps of the target clustering method in the above method. For related content, please refer to the detailed description of the above method, which will not be repeated here.

[0140] The above scheme uses the spatiotemporal information of a small number of image acquisition points of the images to be processed to determine the preset partitions, perform target clustering, determine the clustering information corresponding to each target in the image to be processed, and for the clustering information that needs to be adjusted, determine the adjustment record table, and then adjust the preset partitions to obtain the expected partitions for each image acquisition point, and then partition and classify all the images to be processed according to the expected partitions, effectively determine the optimal target partitions of the images to be processed, and improve the clustering accuracy.

[0141] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0142] The units described as separate components may or may not be physically separate, and 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 may be selected according to actual needs to achieve the purpose of this embodiment.

[0143] In addition, the functional units in various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0144] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, 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. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) or a processor to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0145] The above description is only an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the description and drawings of the present invention, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A target clustering method, characterized in that: include: Acquire spatial information and temporal information of image acquisition points corresponding to at least a portion of the image to be processed, and determine a preset partition of each of the image acquisition points using the spatial information and the temporal information; performing object clustering on at least a portion of the image to be processed to obtain clustering information corresponding to each object in at least a portion of the image to be processed; In response to the clustering information of at least one of the targets being adjusted, obtaining a corresponding adjustment record table, and adjusting the preset partitions using the adjustment record table to determine a desired partition for each of the image acquisition points; All images to be processed are partitioned and classified according to the desired partitions.

2. The method according to claim 1, characterized in that The acquiring of spatial information and temporal information of image acquisition points corresponding to at least a portion of the image to be processed, and determining a preset partition of each image acquisition point using the spatial information and the temporal information, includes: Acquire a plurality of image acquisition points corresponding to at least part of the image to be processed, and acquire spatial information and temporal information of each of the image acquisition points; The plurality of image acquisition points are grouped by utilizing the spatial information and the temporal information to determine a preset partition of each of the image acquisition points, wherein each of the preset partitions includes at least one of the image acquisition points.

3. The method according to claim 2, characterized in that Also includes: Acquire original image information and image acquisition point information of each of the at least part of the images to be processed as point information of each of the images to be processed; In response to the point information and the spatial information being consistent, the preset partition corresponding to the consistent information is used as the current image preset partition of the image to be processed.

4. The method according to claim 1, wherein The performing target clustering on at least a portion of the image to be processed to obtain clustering information corresponding to each target in at least a portion of the image to be processed includes: Acquiring at least one target in at least a portion of the image to be processed; Performing target clustering on the image to be processed containing the target to obtain clustering information corresponding to each target.

5. The method according to claim 1, characterized in that In response to the clustering information of at least one of the targets being adjusted, obtaining a corresponding adjustment record table, and adjusting the preset partitions using the adjustment record table to determine a desired partition for each of the image acquisition points, includes: In response to clustering information of at least one of the targets being adjusted, determining the adjustment record table using adjustment information of the clustering information; The preset partition of each of the image acquisition points is adjusted using the adjustment information in the adjustment record table to determine the desired partition of each of the image acquisition points.

6. The method according to claim 5, characterized in that In response to the clustering information of at least one of the targets being adjusted, determining the adjustment record table based on the adjustment information of the clustering information includes: Obtaining clustering information of each target, and determining adjustment information corresponding to each clustering information; Obtaining an image requiring adjustment corresponding to each of the targets using the adjustment information; The image to be adjusted is adjusted to the correct clustering information of the target, and the adjustment record table is obtained.

7. The method according to claim 5, characterized in that The adjusting the preset partition of each of the image acquisition points by using the adjustment information in the adjustment record table to determine the desired partition of each of the image acquisition points includes: Traversing the adjustment record table of the image to be processed, obtaining the image acquisition point information, the partition information before adjustment, and the partition information after adjustment of each recorded image; and merging the partition information before adjustment and the partition information after adjustment corresponding to the same image acquisition point information into an array to determine an image acquisition point adjustment list; Traversing the image acquisition point adjustment list, obtaining the number of records of the traversed current image acquisition point in the adjustment record table, and obtaining the number of images whose point information in the image to be processed is the current image acquisition point; In response to a ratio of the number of records to the number of images being greater than a preset ratio, obtaining, from the image acquisition point adjustment list, information of the partition where the current image acquisition point is located after adjustment in the array, and determining an information table of the partition where the current image acquisition point is located after adjustment; The adjusted partition information table is traversed to obtain a partition list of the current image acquisition point, and a partition containing images to be processed that is greater than a preset image quantity threshold is used as the desired partition of the current image acquisition point.

8. The method according to claim 7, characterized in that Also includes: Obtaining a record corresponding to the current image acquisition point in the preset partition, and updating the preset partition in the record to the desired partition of the current image acquisition point, wherein the current image acquisition point is the image acquisition point of the desired partition that has been traversed; Obtaining original image information and image acquisition point information of each image to be processed, and determining an image information list; The image information list is traversed to determine the partition information of the updated desired partition of each of the to-be-processed images, and the partition information is used as the image target partition of the corresponding to-be-processed image.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor coupled to the memory, the memory stores at least one computer program, and when the at least one computer program is loaded and executed by the processor, it is used to implement the method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium has at least one program, and when the at least one program is loaded and executed by the processor, it is used to implement the method according to any one of claims 1 to 8.

Images