Target object clustering method, electronic device and computer readable storage medium

Abstract

The application discloses a target object clustering method, an electronic device and a computer readable storage medium. The method comprises the following steps: obtaining feature data corresponding to a plurality of target objects; wherein the target object comprises an object identifier, and the feature data is related to a target position and a target category of the target object; determining a target association type between any two target objects from a plurality of preset association types based on the object identifier and the feature data corresponding to the target object; wherein the plurality of preset association types are mutually distinguished; determining an association degree value between any two target objects based on the target association type and the feature data; obtaining an initial similarity between any two target objects, obtaining a target similarity between any two target objects based on the initial similarity and the corresponding association degree value, and clustering the plurality of target objects by using the target similarity; wherein the target similarity is positively correlated with the association degree value. The above scheme can improve the accuracy of clustering.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a target object clustering method, an electronic device and a computer readable storage medium. BACKGROUND

[0002] With the advent of the data era, a large amount of data needs to be processed and clustered, so as to classify data belonging to the same target object into the same archive and improve the convenience of using data. However, in the prior art, due to the influence of objective factors such as the angle, light and occlusion of the target object collection, the data of the same target object is easily classified into different archives, resulting in poor accuracy of clustering the target object. Therefore, how to improve the accuracy of clustering has become a problem to be solved. SUMMARY

[0003] The technical problem solved by the present application is to provide a target object clustering method, an electronic device and a computer readable storage medium, which can improve the accuracy of clustering.

[0004] To solve the above technical problem, the first aspect of the present application provides a target object clustering method, comprising: obtaining feature data corresponding to a plurality of target objects; wherein the target object comprises an object identifier, and the feature data is related to a target position and a target category of the target object; determining a target association type between any two target objects from a plurality of preset association types based on the object identifier and the feature data corresponding to the target object; wherein the plurality of preset association types are mutually distinguished; determining an association degree value between any two target objects based on the target association type and the feature data; obtaining an initial similarity between any two target objects, obtaining a target similarity between any two target objects based on the initial similarity and the corresponding association degree value, and clustering a plurality of target objects using the target similarity; wherein the target similarity is positively correlated with the association degree value.

[0005] To solve the above technical problem, the second aspect of the present application provides an electronic device, which comprises a memory and a processor coupled to each other, wherein the memory stores program data, and the processor calls the program data to execute the method of the first aspect.

[0006] To solve the above technical problem, the third aspect of the present application provides a computer readable storage medium having program data stored thereon, wherein the program data is executed by a processor to implement the method of the first aspect.

[0007] The scheme is characterized in that: the plurality of target objects and the feature data corresponding to the target objects are obtained, wherein the target objects comprise object identifiers, the feature data is related to target positions and target categories corresponding to the target objects; one of the plurality of preset association types is selected as a target association type between any two target objects based on the object identifiers and the feature data corresponding to the target objects, wherein different preset association types are distinguished from each other; an association degree value between any two target objects is obtained based on the target association type and the feature data corresponding to the target objects; an initial similarity between any two target objects is obtained; and a target similarity between any two target objects is obtained based on the initial similarity and the corresponding association degree value, wherein the target similarity is positively related to the association degree value, so that the association degree value plays a correcting role on the initial similarity, thereby obtaining a more accurate target similarity; and the plurality of target objects are clustered by using the target similarity, thereby improving the accuracy of clustering. BRIEF DESCRIPTION OF DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor. Among them:

[0009] Figure 1 is a flowchart of an embodiment of a target object clustering method of the present application;

[0010] Figure 2 is a flowchart of another embodiment of a target object clustering method of the present application;

[0011] Figure 3 is a structural schematic diagram of an embodiment of an electronic device of the present application;

[0012] Figure 4 is a structural schematic diagram of an embodiment of a computer readable storage medium of the present application. DETAILED DESCRIPTION

[0013] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0014] If the technical solution of the present application involves personal information, the product applying the technical solution of the present application has been explicitly informed of the personal information processing rules before processing the personal information and has obtained the personal independent consent. If the technical solution of the present application involves sensitive personal information, the product applying the technical solution of the present application has obtained the personal independent consent before processing the sensitive personal information and at the same time meets the requirement of "explicit consent". For example, at the personal information collection device such as camera, a clear and prominent sign is set to inform that the personal information collection range has been entered and the personal information will be collected. If the person voluntarily enters the collection range, it is regarded as consent to collect the personal information. Or, on the device for processing personal information, the personal information processing rules are informed by using obvious signs / information, and the personal authorization is obtained by means of pop-up information or asking the person to upload his / her personal information. The personal information processing rules can include personal information processor, personal information processing purpose, processing method and personal information type.

[0015] The terms "system" and "network" are often used interchangeably herein. The term "and / or" herein is only a description of the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can represent the three cases of A alone, A and B together, and B alone. In addition, the character " / " herein generally represents an "or" relationship between the front and rear associated objects. In addition, "multiple" herein means two or more than two.

[0016] The target object archiving method provided by the present application is used for archiving data corresponding to a target object, wherein the data corresponding to the target object at least includes images or videos. For example, the target object is a person, and the data corresponding to the target object is a portrait obtained from the images or videos, so as to archive the portrait. Of course, the target object can correspond to multiple target categories, and the present application will not be repeated here. The execution subject of the target object archiving method provided by the present application is a processor capable of calling data.

[0017] Please refer to Figure 1 , Figure 1 is a flowchart of an embodiment of the target object archiving method of the present application. The method comprises:

[0018] S101: Obtain feature data corresponding to a plurality of target objects, wherein the target object includes an object identifier, and the feature data is related to a target position and a target category of the target object.

[0019] Specifically, a plurality of target objects and feature data corresponding to the target objects are obtained, wherein the target object includes an object identifier, and the feature data is obtained based on a target position and a target category corresponding to the target object.

[0020] In an application mode, target positions and target categories corresponding to a plurality of target objects are obtained, trajectory information of the target objects is determined based on the target positions corresponding to the target objects at different time nodes, attribute information corresponding to the target objects is determined based on the target categories corresponding to the target objects, and feature data corresponding to the target objects is obtained based on the trajectory information and the attribute information corresponding to the target objects.

[0021] In another application mode, target positions and target categories corresponding to a plurality of target objects are obtained, relationship information between any two target objects is determined based on the target positions and the target categories corresponding to the target objects, trajectory information formed by the target objects within a preset time period is determined based on the target positions corresponding to the target objects, attribute information corresponding to the target objects and matching the target categories is obtained, and feature data corresponding to the target objects is obtained based on the trajectory information, the attribute information, and the relationship information corresponding to the target objects.

[0022] It should be noted that at least part of the target objects have relationship information with at least one other target object. When two target objects belong to the same target category, for example, both target objects belong to the category of people, the relationship information between the target objects can be one of a plurality of preset interpersonal relationships, and the preset interpersonal relationships at least include direct relatives, collateral relatives, friends, colleagues, and teachers and students. When the target objects belong to different target categories, for example, one target object is a person and the other target object is an article, the relationship information between the target objects can be one of a plurality of preset binding relationships, and the preset binding relationships at least include a subordinate relationship, a temporary occupation relationship, and a long-term occupation relationship.

[0023] Further, target objects of different categories correspond to object identifiers matching the target categories. The target categories of the target objects at least include people, articles, and data sets. The object identifier corresponding to the people is an identity identifier, the object identifier corresponding to the articles is an article identifier, and the object identifier corresponding to the data sets is a data identifier.

[0024] In an application scenario, the identity identifier corresponds to the identity card number of the person, the identity card number serves as the object identifier of the person, the article identifier corresponds to the identification code of the article, and the identification code serves as the article identifier of the article. For example, a mobile phone corresponds to a mobile phone number or a network access permit number, a vehicle corresponds to a license plate number or a vehicle frame number, a computer corresponds to a physical address, and a data identifier corresponds to a file number of a data set, and the file number serves as the data identifier of the data set. The file number can be a combination of any letters and / or numbers.

[0025] S102: Based on the object identifiers and the feature data corresponding to the target objects, a target association type between any two target objects is determined from a plurality of preset association types, wherein the plurality of preset association types are mutually distinguished.

[0026] Specifically, based on the object identifier and the feature data corresponding to the target objects, one of the plurality of preset association types is selected as the target association type between any two target objects, wherein the different preset association types are distinguished from each other.

[0027] In an application mode, the plurality of preset association types correspond to respective association degrees, so that the plurality of preset association types are distinguished from each other. When the target objects correspond to the same object identifier, the preset association type with the highest association degree is selected from the preset association types as the target association type between the target objects. When the target objects do not correspond to the same object identifier, the relationship information between the target objects is determined based on the feature data corresponding to the target objects, and one of the preset association types is selected as the target association type between the target objects based on the relationship information.

[0028] In another application mode, the preset association type corresponds to a first association type and a second association type, and the association degree of the first association type is greater than the association degree of the second association type. The relationship information between the target objects is determined based on the feature data corresponding to the target objects, the object identifier corresponding to the other target object corresponding to the relationship information is obtained based on the relationship information, and all object identifiers corresponding to each target object are obtained. When any two target objects correspond to at least one group of same object identifiers, the first association type is taken as the target association type between the target objects. When any two target objects do not correspond to any group of same object identifiers, the second association type is taken as the target association type between the target objects.

[0029] S103: Determine the association degree value between any two target objects based on the target association type and the feature data.

[0030] Specifically, based on the target association type and the feature data corresponding to the target objects, the association degree value between any two target objects is obtained.

[0031] In an application mode, each target association type corresponds to a respective initial degree value, and the initial degree value is corrected based on the feature data corresponding to the target objects, so that the association degree value between any two target objects is obtained, so that a more accurate association degree value is obtained based on the initial degree value.

[0032] In another application mode, in response to the target association type being a preset association type with the highest association degree, a preset value is set between the corresponding target objects to obtain the association degree value, and in response to the target association type being other preset association types different from the preset association type with the highest association degree, a reference value is set between the corresponding target objects based on the feature data corresponding to the target objects to obtain the association degree value, wherein the reference value is less than the preset value, so that the target objects with a higher association degree are set with the preset value and the target objects with a lower association degree are set with the reference value, and the rationality of the association degree value is improved.

[0033] S104: Obtain the initial similarity between any two target objects, obtain the target similarity between any two target objects based on the initial similarity and the corresponding association degree value, and use the target similarity to cluster the plurality of target objects, wherein the target similarity is positively correlated with the association degree value.

[0034] Specifically, based on the initial similarity and the corresponding association degree value, the target similarity between any two target objects is obtained, wherein the target similarity is positively correlated with the association degree value, so that the association degree value plays a correcting role on the initial similarity, so that a more accurate target similarity is obtained, and the plurality of target objects are clustered using the target similarity, so that the accuracy of clustering is improved.

[0035] In an application mode, the initial similarity between any two target objects is obtained, the initial similarity is corrected using the association degree value to obtain the target similarity between any two target objects, when the association degree value is greater than a degree value threshold, the target similarity is greater than the initial similarity, and when the association degree value is less than the degree value threshold, the target similarity is less than the initial similarity, wherein the initial similarity is obtained based on the video or image corresponding to the two target objects.

[0036] In another application mode, the images corresponding to any two target objects are obtained, the initial similarity between any two target objects is determined based on the similarity of the target objects on the image, an adjustment coefficient is determined based on the initial similarity, a corresponding adjustment variable is determined based on the association degree value, and the initial similarity is adjusted using the product of the adjustment coefficient and the adjustment variable to obtain the target similarity, wherein the adjustment variable is positively correlated with the association degree value.

[0037] Further, the target objects with a target similarity exceeding a similarity threshold are classified into the same data set to obtain the clustering result corresponding to the plurality of target objects.

[0038] The scheme obtains a plurality of target objects and corresponding feature data of the target objects, wherein the target objects include object identifiers, the feature data is related to target positions and target categories corresponding to the target objects, an object identifier and feature data corresponding to a target object are used to select a preset association type from a plurality of preset association types as a target association type between any two target objects, wherein different preset association types are distinguished from each other, an association degree value between any two target objects is obtained based on the target association type and the feature data corresponding to the target objects, an initial similarity between any two target objects is obtained, and a target similarity between any two target objects is obtained based on the initial similarity and the corresponding association degree value, wherein the target similarity is positively related to the association degree value, so that the association degree value plays a correcting role on the initial similarity, thereby obtaining a more accurate target similarity, and the target similarity is used to cluster the plurality of target objects, thereby improving the accuracy of clustering.

[0039] Referring to Figure 2 , Figure 2 is a flowchart of another embodiment of the target object clustering method, which includes the following steps.

[0040] S201: Obtain target positions and target categories corresponding to a plurality of target objects, wherein a plurality of target positions corresponding to the target objects matched by the same object identifier form trajectory information of the target objects, each target object corresponds to attribute information matched with the target category, and relationship information corresponding to at least part of the target objects is related to the target positions and the target categories.

[0041] Specifically, the target positions and the target categories corresponding to the plurality of target objects are obtained, wherein each target object corresponds to a respective object identifier, the trajectory information corresponding to the target objects is determined based on the target positions corresponding to the target objects, and the attribute information matched with the target category is determined based on the target categories corresponding to the target objects.

[0042] Further, the relationship information between at least part of the target objects is determined based on the target positions and the target categories corresponding to the plurality of target objects, thereby deriving trajectory information, attribute information, and relationship information with more prominent features based on the target positions and the target categories corresponding to the target objects.

[0043] In an application mode, the trajectory information includes target positions and time stamps corresponding to the target objects, the attribute information includes attribute features matched with the target categories of the target objects, and the relationship information includes at least one preset relationship category corresponding to any two target objects and a corresponding confidence degree.

[0044] Specifically, the trajectory information includes target positions of a same object identifier corresponding to a target object at different timestamps, such that a line connecting the target positions at different timestamps forms the trajectory information, and the attribute information is highly matched with a target category of the target object, so as to reflect the attribute feature of the target object.

[0045] Further, at least one preset relationship category exists between any two target objects with relationship information, and the preset relationship category corresponds to a confidence degree. That is, at least one of the preset relationship categories is selected between the target objects, and the corresponding preset relationship category includes the confidence degree to reflect the credibility of the preset relationship category.

[0046] In an application scenario, the trajectory information of the target object includes a trajectory generated by the spatio-temporal movement of the target object, and a data format of the trajectory information of the target object is represented as (Id, GPS_X, GPS_Y, T). Wherein, GPS_X and GPS_Y represent longitude and latitude information of the movement of the target object, T represents a timestamp, and the longitude and latitude information correspond to the position of the target object. The attribute information of the target object includes attribute features matched with the target category of the target object. For example, the target object is a mobile phone, and the associated attribute information at least includes an identity of a household and a model of the mobile phone; the target object is a person, and the associated attribute information at least includes height, age, and birthplace.

[0047] Further, the relationship information between the target objects at least includes a relationship with spatio-temporal behavior and a relationship with a family and friend category, wherein the relationship with the same spatio-temporal behavior corresponds to a preset relationship category at least including the same time period appearing or disappearing or the number of times of appearing or disappearing at the same place exceeding a threshold, and the relationship with the family and friend category corresponds to a preset relationship category at least including mother-child, father-child, mother-daughter, father-daughter, husband-wife, and colleague. It can be understood that the relationship with the same spatio-temporal behavior is determined based on the trajectory information, the relationship with the family and friend category is determined based on the attribute information, and when a relationship table of the target objects is obtained, the confidence degree of the preset relationship category between the corresponding target objects is the maximum value, wherein the relationship table includes the relationship category between the target objects obtained in advance.

[0048] In a specific application scenario, a specific representation form of the relationship information can be (Id1, Id2, R_T, R_V). Wherein, R_T is an enumeration value, representing different relationship types between the target objects, wherein each enumeration value corresponds to a preset relationship category. R_V is a floating point value, representing the confidence degree of the relationship type R_T between the target objects, used to represent the probability of the existence of the relationship. Wherein, the confidence degree is between 0 and 1, and the larger the value is, the higher the confidence degree is.

[0049] It should be noted that the relationship between the target objects is directionless, that is, (Id1, Id2, R_T, R_V) and (Id2, Id1, R_T, R_V) represent the same relationship information between two target objects. Therefore, for the same relationship information, only one direction of relationship information needs to be stored when storing.

[0050] S202: Based on the different types of information respectively matched with the feature construction mode, the trajectory information, the attribute information and the relationship information are constructed respectively to obtain the feature data composed of multiple feature information.

[0051] Specifically, different types of information correspond to respective feature construction modes, thereby adapting to different types of information. The trajectory information, the attribute information and the relationship information are respectively constructed by using the respective feature construction modes to obtain the feature data composed of multiple feature information.

[0052] Optionally, before constructing the feature information, the trajectory information, the attribute information and the relationship information are further subjected to data processing, wherein the data processing includes outlier rejection, duplicate value de-duplication and missing value filling, so as to obtain processed information, thereby improving the accuracy of the information.

[0053] In an application scenario, the feature construction mode of the trajectory information includes data aggregation with object identification as an index, and mathematical statistical features are used to represent corresponding trajectory feature information, wherein the mathematical statistical features include median, mean, maximum or minimum. The feature construction mode of the attribute information includes data segmentation according to the type of attribute features, for example, age data is divided into multiple age ranges. The feature construction mode of the relationship information includes extracting a preset relationship category with a confidence exceeding a confidence threshold.

[0054] S203: Based on the object identification corresponding to the target object and the feature data, a target association type between any two target objects is determined from multiple preset association types, wherein the multiple preset association types are mutually distinguished.

[0055] Specifically, the multiple target objects correspond to multiple target categories, and the target objects of different target categories correspond to object identifications matched with the target categories.

[0056] In an application mode, the target object is specifically a companion object, which is an entity object with a unique identification and has a space-time moving ability. The target category of the target object can be a person, and the person is represented by an identity identification as a unique identification. The target category of the target object can be an object, and the object is represented by an object unique identification. The target category of the target object can be a profile of a file, and the profile is represented by a profile identification as a unique identification.

[0057] Further, based on the object identifiers corresponding to the target objects and the feature data, a target association type between any two target objects is determined from a plurality of preset association types, including: based on the relationship information included in the feature data corresponding to the target objects, determining the object identifiers corresponding to the target objects and the object identifiers corresponding to other target objects corresponding to the relationship information of the target objects; based on all object identifiers corresponding to the target objects, determining the target association type between any two target objects from a plurality of preset association types; wherein the association degrees corresponding to the plurality of preset association types are mutually distinguished, and the target association type is related to the number of the same object identifiers included between any two target objects.

[0058] Specifically, if based on the feature data corresponding to the current target object, relationship information corresponding to other target objects can be obtained, the object identifier corresponding to the current target object and the object identifier of the other target object having relationship information with the current target object are obtained.

[0059] In an application scenario, a person and a mobile phone have an association relationship, the object identifier corresponding to the person is an ID card number, and the object identifier corresponding to the mobile phone is a mobile phone number. The object identifiers corresponding to the person and the mobile phone both include the ID card number and the mobile phone number.

[0060] Further, based on all object identifiers corresponding to the target objects, the number of the same object identifiers included between any two target objects is determined, so that based on the number of the same object identifiers included between the two target objects, the target association type between any two target objects is determined from a plurality of preset association types, wherein the association degrees corresponding to the plurality of preset association types are mutually distinguished, and the target association type is related to the number of the same object identifiers included between any two target objects.

[0061] In an application scenario, the preset association types include at least three association types of association degrees, when the target objects correspond to more than the upper limit of the number of the same object identifiers, the preset association type with the highest association degree is selected as the target association type for the corresponding target objects, when the target objects correspond to the same object identifiers between the number lower limit and the number upper limit, the preset association type corresponding to the association degree matching the number is selected as the target association type for the corresponding target objects, and when the target objects correspond to less than the number of the same object identifiers The lower limit, the preset association type with the lowest association degree is selected as the target association type for the corresponding target objects, so as to configure more accurate target association types for the target objects.

[0062] In another application scenario, the preset association types include a first association type and a second association type, where the association degree of the first association type is higher than that of the second association type; and the target association type between any two target objects is determined from the plurality of preset association types based on all object identifiers corresponding to the target objects, including: in response to the number of groups of target objects corresponding to the same object identifier being greater than or equal to a quantity threshold, setting the target association type of the corresponding target objects as the first association type; and in response to the number of groups of target objects corresponding to the same object identifier being less than the quantity threshold, setting the target association type of the corresponding target objects as the second association type.

[0063] Specifically, the preset association types include a first association type and a second association type with mutually different association degrees, where the association degree of the first association type is higher than that of the second association type, that is, the first association type is a strong association relationship and the second association type is a weak association relationship.

[0064] Further, when the number of groups of target objects corresponding to the same object identifier is greater than or equal to the quantity threshold, the target association type of the corresponding target objects is set as the first association type, and when the number of groups of target objects corresponding to the same object identifier is less than the quantity threshold, the target association type of the corresponding target objects is set as the second association type, so as to improve the efficiency of obtaining the target association type.

[0065] In a specific application scenario, the quantity threshold is 1, the current target object is a person and the current target object has relationship information with a mobile phone, the object identifier corresponding to the current target object includes an ID number and a mobile phone number, and other target objects different from the current target object are traversed. If any target object has the same ID number or the same mobile phone number as the current target object, the target association type between the corresponding target object and the current target object is the first association type, and if any target object does not have the same ID number or the same mobile phone number as the current target object, the target association type between the corresponding target object and the current target object is the second association type.

[0066] S204: Determine the association degree value between any two target objects based on the target association type and the feature data.

[0067] Specifically, the association degree value between any two target objects is obtained based on the target association type and the feature data corresponding to the target objects.

[0068] In an application mode, in response to the target association type between the two target objects being a specified association type, the association degree value between the corresponding two target objects is determined as a preset value; wherein the specified association type is a preset association type with the highest association degree; in response to the target association type between the two target objects being a preset association type other than the specified association type, the association degree value between the corresponding two target objects is determined based on the feature data corresponding to the target objects.

[0069] Specifically, when the target association type between the two target objects is the specified association type, the association degree value between the corresponding two target objects is determined as a preset value, wherein the preset value can be the maximum value of the association degree value.

[0070] Further, when the target association type between the two target objects is a preset association type other than the specified association type, the association degree between the target objects is analyzed based on the feature data corresponding to the target objects, and the association degree value between the corresponding two target objects is obtained.

[0071] In an application scenario, the association strength value is valued as 0-1, when the target association type between the two target objects is the specified association type, the association degree value between the corresponding two target objects is determined as a preset value, wherein the preset value is 1, when the target association type between the two target objects is a preset association type other than the specified association type, the association degree value between the target objects is determined based on the feature data of the target objects, wherein the trajectory information, attribute information and relationship information of the target objects are determined based on the target position and target category of the target objects, the initial association strength value between the target objects is determined by using the similarity of the trajectory information, and the initial association strength value is corrected by using the attribute information and relationship information, wherein the similarity of the trajectory information can be obtained based on the DTW (Dynamic Time Warping) or LCSS (Longest Common Sub-Sequence) algorithm, each attribute information and relationship information corresponds to a preset correction value, so as to fuse the correction value based on the initial similarity, and obtain the association strength value between the two target objects.

[0072] S205: Based on the images / videos corresponding to any two target objects, the initial similarity between any two target objects is determined.

[0073] Specifically, based on the images or videos corresponding to any two target objects, the initial similarity between the target objects is determined, wherein any target object can be a target object already archived in the archive set or a target object not archived outside the archive set.

[0074] S206: Correct the initial similarity degree by using the correlation degree value to obtain the target similarity degree between any two target objects.

[0075] Specifically, the initial similarity degree is corrected by using the correlation degree value, and the corrected similarity degree is taken as the target similarity degree between any two target objects, thereby improving the accuracy of the target similarity degree, wherein the target similarity degree is positively correlated with the correlation degree value.

[0076] In an application mode, a backoff constant is obtained based on the initial similarity degree, a correction constant is determined based on the backoff constant; wherein the initial similarity degree is between 0-1, the backoff constant is less than the initial similarity degree, and the correction constant is the difference between 1 and the backoff constant; an adjustment coefficient corresponding to the correction constant is determined based on the initial similarity degree, an adjustment variable corresponding to the correction constant is determined based on the correlation degree value, and the correction constant is adjusted by using the product of the adjustment coefficient and the adjustment variable to obtain a correction value; wherein the adjustment variable is positively correlated with the correlation degree value; the difference between the initial similarity degree and the backoff constant is obtained, and the sum of the difference and the correction value is taken as the target similarity degree.

[0077] Specifically, the initial similarity degree is between 0-1, a backoff constant is obtained based on the initial similarity degree, wherein the backoff constant is less than the initial similarity degree, the difference between 1 and the backoff constant is taken as a correction constant, an adjustment coefficient corresponding to the correction constant is determined based on the initial similarity degree, and an adjustment variable corresponding to the correction constant is determined based on the correlation degree value, wherein the adjustment variable is positively correlated with the correlation degree value, thereby adjusting the correction constant by using the product of the adjustment coefficient and the adjustment variable to obtain a correction value. Therefore, the initial similarity degree and the correlation strength value are both used as variables to correct the initial similarity degree, thereby improving the accuracy of the target similarity degree obtained finally, the difference between the initial similarity degree and the backoff constant is obtained, and the sum of the difference and the correction value is taken as the target similarity degree. The above process is represented by the following formula:

[0078] α′=α-δ+(1-δ)e -α1ogβ-(1-α)log(1-β) (1)

[0079] Wherein, α' is the target similarity degree, α is the initial similarity degree, δ is the backoff constant, and 0<δ<α, (1-δ) is the correction constant, α and (1-α) are respectively taken as the adjustment coefficient, and logβ and log(1-β) are respectively taken as the adjustment variable, wherein 0.5<β<1.

[0080] S207: Based on the target similarity degree and the similarity threshold value corresponding to the target similarity degree, the multiple target objects are clustered.

[0081] Specifically, the target objects whose target similarity degree exceeds the similarity threshold value are classified into the same data set to obtain the clustering result corresponding to the multiple target objects, thereby improving the accuracy of the clustering.

[0082] In the embodiment, the correlation degree value between any two target objects is obtained based on the target correlation type and the feature data corresponding to the target objects, the initial similarity is corrected by using the correlation degree value, and the corrected similarity is taken as the target similarity between any two target objects, so as to improve the accuracy of the target similarity, and then the target objects with the target similarity exceeding the similarity threshold are classified into the same data set, the clustering result corresponding to the target objects is obtained, and the accuracy of the clustering is improved.

[0083] Please refer to Figure 3 , Figure 3 is a structural schematic diagram of an embodiment of an electronic device of the present application. The electronic device 30 includes a memory 301 and a processor 302 coupled with each other. The memory 301 stores program data (not shown in the figure). The processor 302 invokes the program data to implement the attribute classification method based on video frames in any of the above embodiments. For related content, please refer to the detailed description of the method embodiment, which will not be repeated here.

[0084] Please refer to Figure 4 , Figure 4 is a structural schematic diagram of an embodiment of a computer readable storage medium of the present application. The computer readable storage medium 40 stores program data 400. The program data 400 is executed by a processor to implement the attribute classification method based on video frames in any of the above embodiments. For related content, please refer to the detailed description of the method embodiment, which will not be repeated here.

[0085] It should be noted that the units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or they can be distributed on multiple network units. According to actual needs, part or all of the units can be selected to achieve the purpose of the present embodiment.

[0086] In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0087] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the present application, essentially or in other words, the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform all or part of the steps of the methods in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0088] The above is only the embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent flow transformation made by using the content of the specification and drawings, or directly or indirectly applied to other related technical fields, is also included in the patent protection scope of the present application.

Claims

1. A method for clustering target objects, characterized in that, The method includes: Obtain feature data corresponding to multiple target objects; wherein, the target object includes an object identifier, and the feature data is related to the target location and target category of the target object; Based on the object identifier and feature data corresponding to the target object, a target association type between any two target objects is determined from a plurality of preset association types; wherein, the plurality of preset association types are distinct from each other; Based on the target association type and the feature data, determine the degree of association between any two target objects; The process involves obtaining an initial similarity between any two target objects, obtaining a target similarity between any two target objects based on the initial similarity and the corresponding association value, and then using the target similarity to cluster multiple target objects. Specifically, this includes: determining the initial similarity between any two target objects based on their corresponding images / videos; correcting the initial similarity using the association value to obtain the target similarity between any two target objects; and clustering multiple target objects based on the target similarity and a similarity threshold corresponding to the target similarity. The target similarity is positively correlated with the association value.

2. The target object aggregation method according to claim 1, characterized in that, The process of obtaining feature data corresponding to multiple target objects includes: Obtain target locations and target categories corresponding to multiple target objects; wherein, multiple target locations corresponding to target objects matched by the same object identifier form the trajectory information of the target objects, each target object corresponds to attribute information that matches the target category, and at least some of the target objects correspond to relationship information, the relationship information being related to the target location and the target category; Based on the feature construction method of matching different types of information, feature information corresponding to the trajectory information, the attribute information and the relationship information are constructed respectively, and feature data composed of multiple feature information is obtained.

3. The target object aggregation method according to claim 2, characterized in that, The trajectory information includes the target location and timestamp corresponding to the target object, the attribute information includes attribute features that match the target category of the target object, and the relationship information includes at least one preset relationship category and its corresponding confidence level between any two target objects.

4. The target object aggregation method according to claim 2, characterized in that, The multiple target objects correspond to the multiple target categories, and the target objects of different target categories correspond to object identifiers that match the target category; The step of determining the target association type between any two target objects from a plurality of preset association types based on the object identifier corresponding to the target object and the feature data includes: Based on the relational information included in the feature data corresponding to the target object, determine the object identifier corresponding to the target object, and the object identifiers corresponding to other target objects that have the relational information corresponding to the target object; Based on all the object identifiers corresponding to the target object, a target association type between any two target objects is determined from a plurality of preset association types; wherein, the degree of association corresponding to the plurality of preset association types is different from each other, and the target association type is related to the number of identical object identifiers included between any two target objects.

5. The target object aggregation method according to claim 4, characterized in that, The preset association types include a first association type and a second association type, wherein the degree of association of the first association type is higher than the degree of association of the second association type; The step of determining the target association type between any two target objects from a plurality of preset association types based on all object identifiers corresponding to the target object includes: In response to the number of groups corresponding to the same object identifier of the target object being greater than or equal to a quantity threshold, the target association type of the corresponding target object is set to the first association type; In response to the fact that the number of groups corresponding to the same object identifier for the target object is less than a quantity threshold, the target association type of the corresponding target object is set to the second association type.

6. The target object aggregation method according to claim 1, characterized in that, The step of determining the degree of association between any two target objects based on the target association type and the feature data includes: In response to the target association type between two target objects being a specified association type, the degree of association between the two target objects is determined to be a preset value; wherein, the specified association type is the preset association type with the highest degree of association; In response to the target association type between the two target objects being a preset association type that is different from the specified association type, the degree of association between the two target objects is determined based on the feature data corresponding to the target objects.

7. The target object aggregation method according to claim 1, characterized in that, The step of correcting the initial similarity using the correlation degree value to obtain the target similarity between any two target objects includes: A backoff constant is obtained based on the initial similarity, and a correction constant is determined based on the backoff constant; wherein, the initial similarity is between 0 and 1, the backoff constant is less than the initial similarity, and the correction constant is the difference between 1 and the backoff constant; Based on the initial similarity, an adjustment coefficient corresponding to the correction constant is determined; based on the correlation value, an adjustment variable corresponding to the correction constant is determined; and the correction constant is adjusted using the product of the adjustment coefficient and the adjustment variable to obtain a correction value; wherein, the adjustment variable is positively correlated with the correlation value. The difference between the initial similarity and the backoff constant is obtained, and the sum of the difference and the correction value is taken as the target similarity.

8. An electronic device, characterized in that, include: A memory and a processor are coupled to each other, wherein the memory stores program data, and the processor invokes the program data to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium storing program data thereon, characterized in that, When the program data is executed by the processor, the method as described in any one of claims 1-7 is implemented.

Images