A method, apparatus and medium for merging archives

Abstract

This application proposes a file merging method, comprising: acquiring files to be merged, each file including a cover image and multiple face images; searching a database based on the files to be merged to obtain pre-merged files containing multiple face images, and constructing candidate file pairs from the files to be merged and the pre-merged files; acquiring multi-dimensional feature information of the files to be merged and the pre-merged files, as well as interaction feature information of the candidate file pairs; inputting the multi-dimensional feature information and the interaction feature information into a trained machine learning model, wherein the machine learning model outputs a confidence score indicating that two files in the candidate file pairs belong to the same target; and determining whether the candidate file pairs should be merged into a merged file based on the confidence score. The file merging method of this application integrates more multi-dimensional feature information, resulting in higher accuracy and merging rate, while also achieving deduplication between different merged files.

Description

Technical Field

[0001] This application relates to the technical field of image processing, specifically to a file merging method, apparatus, and medium. Background Technology

[0002] With the development of artificial intelligence and the widespread use of cameras, facial recognition technology is becoming increasingly mature, and high-definition video surveillance coverage in the public security field is expanding, making facial image clustering and archiving possible. However, due to various external factors at different times, the structured information of facial images of the same target has low feature similarity, making it difficult to identify them as the same target. This results in multiple files for the same person in the facial archive, degrading the archiving effect and affecting the effectiveness of subsequent techniques and tactics based on facial image clustering. Therefore, it is necessary to merge and deduplicate facial archives.

[0003] Existing technical solutions are generally based on feature similarity, merging files that meet or exceed a preset threshold by calculating the facial feature similarity between two files. Current technologies for calculating facial feature similarity between files typically employ two methods: one uses a representative image (also called a cover image) to represent each file, and calculates the similarity between these representative images to represent the file's overall similarity; the other uses class central features to represent each file, and calculates the similarity between these class central features to represent the average similarity between the files. Additionally, some technical solutions combine both methods. However, these solutions only utilize information from the dimension of image feature similarity and do not fully consider other influencing factors, resulting in a relatively low merging rate. Summary of the Invention

[0004] To address the technical problem of low file merging rates in existing technologies, this application proposes a file merging method, apparatus, and medium.

[0005] According to the first aspect of this application, a method for merging archives is proposed, comprising:

[0006] Obtain the files to be merged, each file including a cover image and multiple face images;

[0007] The archive database is searched according to the archives to be merged to obtain a pre-merged archive containing multiple face images, and the archives to be merged and the pre-merged archives are constructed into a candidate archive pair;

[0008] Obtain multi-dimensional feature information of the files to be merged and the files to be merged, as well as the interaction feature information of the candidate file pairs;

[0009] The multi-dimensional feature information and the interaction feature information are input into the trained machine learning model, and the machine learning model outputs the confidence that the two files in the candidate file pair belong to the same target;

[0010] Based on the confidence level, determine whether the candidate file pairs should be merged into a merged file.

[0011] Preferably, the step of retrieving the archive database based on the archives to be merged includes:

[0012] The archive database is retrieved based on the archive features of the archives to be merged. The archive features of the archives to be merged are cover image characteristics or class center features, wherein the class center features are used to characterize the average vector of feature vectors of multiple face images in the archives to be merged.

[0013] Preferably, the step of retrieving a pre-merge file containing multiple face images from the archive to be merged, and constructing a candidate archive pair from the archive to be merged and the pre-merge file, includes:

[0014] The archive database is searched based on the archives to be merged to obtain one or more pre-merged archives with different similarities;

[0015] The pre-merged files whose similarity reaches a first preset threshold in the search results are respectively constructed with the files to be merged to form the candidate file pairs.

[0016] Preferably, the step of retrieving a pre-merged archive containing multiple face images from the archive to be merged, and constructing a candidate archive pair from the archive to be merged and the pre-merged archive, further includes:

[0017] Analyze the search results;

[0018] If the number of pre-merged files whose similarity reaches a second preset threshold is greater than a third preset threshold, then only the pre-merged files whose similarity reaches the second preset threshold are selected and combined with the file to be merged to form the candidate file pair; otherwise, only the pre-merged file with the highest similarity in the search results is selected and combined with the file to be merged to form the candidate file pair.

[0019] Preferably, the multi-dimensional feature information of the files to be merged and the files to be merged includes:

[0020] The facial images in the archive to be merged are categorized by age, gender, whether a mask is worn, average quality score, cohesion, and number of images. The cohesion is used to characterize the average similarity of the feature vectors of any two facial images in the archive.

[0021] The number of pre-merged files retrieved, and the number of pre-merged files with a similarity greater than a fourth preset threshold.

[0022] Preferably, the interaction feature information of the candidate file pair includes:

[0023] The candidate file pair includes the feature similarity between two files, the cohesion after merging the two files, the reduction in cohesion, and the average quality score of the face image.

[0024] Preferably, the expression for the cohesion of the two files in the candidate file pair after merging is specifically as follows:

[0025]

[0026] in,

[0027]

[0028] x = (x1, x2, ..., x n ) represents the feature vector of the facial image in the archive, s a and s b The cohesion of the two files, n a and n b , where represents the number of face images in the two files, and sim represents the feature similarity between the two files;

[0029] The expression for the reduction in cohesion after merging two files in the candidate file pair is as follows:

[0030] s ab -0.5×(s a +s b ).

[0031] Preferably, determining whether the candidate file pairs should be merged into a merged file based on the confidence level includes:

[0032] Filter out all candidate file pairs whose confidence level reaches the fifth preset threshold, and construct a graph network from the files to be merged and the files to be merged in all candidate file pairs;

[0033] The pruning sequence is constructed by sorting all the files to be merged in the graph network in descending order of the number of face images.

[0034] According to the pruning sequence, the file to be merged and all the pre-merged files associated with it are merged into the merged file, and removed from the graph network after each merge.

[0035] According to a second aspect of this application, a file merging device is proposed, comprising:

[0036] The acquisition module is configured to acquire files to be merged, each of which includes a cover image and multiple face images;

[0037] The retrieval module is configured to retrieve the archive database based on the archive to be merged, obtain the pre-merged archive containing multiple face images, and construct the archive to be merged and the pre-merged archive into a candidate archive pair;

[0038] The feature extraction module is configured to acquire multi-dimensional feature information of the files to be merged and the files to be merged, as well as the interaction feature information of the candidate file pairs;

[0039] The prediction module is configured to input the multi-dimensional feature information and the interaction feature information into a trained machine learning model, and the machine learning model outputs the confidence that the two files in the candidate file pair belong to the same target;

[0040] The strategy module is configured to determine whether the candidate file pairs should be merged into a merged file based on the confidence level.

[0041] According to a third aspect of this application, a computer-readable storage medium is provided that stores a computer program, which, when executed by a processor, implements the file merging method as described in the first aspect of this application.

[0042] This application proposes a method, apparatus, and medium for merging archives. By acquiring multi-dimensional feature information of the archives to be merged and the pre-merging archives, as well as the interaction feature information of candidate archive pairs, a machine learning model is then used to determine the confidence level that two archives belong to the same target. This results in higher accuracy and a higher merging rate, effectively mitigating the phenomenon of multiple archives with the same target in the archive repository. Furthermore, after selecting candidate archive pairs with confidence levels meeting a preset threshold through the machine learning model, all candidate archive pairs are constructed into a graph network, and a pruning strategy is used to perform pruning operations, thereby achieving deduplication while merging archives, resulting in an even higher merging rate. Attached Figure Description

[0043] The accompanying drawings are included to provide a further understanding of the embodiments and are incorporated in and constitute a part of this specification. The drawings illustrate embodiments and, together with the description, serve to explain the principles of this application. Other embodiments and many anticipated advantages of these embodiments will be readily recognized as they become better understood through reference to the following detailed description. Elements in the drawings are not necessarily to scale. The same reference numerals refer to corresponding similar parts.

[0044] Figure 1 This is a flowchart of a file merging method according to an embodiment of this application;

[0045] Figure 2This is a schematic diagram of a pruning strategy according to a specific embodiment of this application;

[0046] Figure 3 This is a structural block diagram of a file merging device according to an embodiment of this application. Detailed Implementation

[0047] The features and exemplary embodiments of various aspects of this application will now be described in detail. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only configured to explain this application and are not configured to limit this application. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples of this application.

[0048] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0049] According to the first aspect of this application, a method for merging archives is proposed. Figure 1 A flowchart of a file merging method according to an embodiment of this application is shown, such as... Figure 1 As shown, the method includes the following steps:

[0050] S101: Obtain the files to be merged. Each file to be merged includes a cover image and multiple face images.

[0051] In a specific embodiment, the cover image is the most representative image (e.g., the one with the highest image quality score) in each file to be merged. The criteria for obtaining the files to be merged are as follows: 1. Prioritize newly generated files; 2. Prioritize files with a large number of face images; 3. Prioritize files that have been selected the longest since they were last selected.

[0052] S102: Search the archive database based on the archives to be merged to obtain a pre-merged archive containing multiple face images, and construct a candidate archive pair from the archives to be merged and the pre-merged archives.

[0053] In a specific embodiment, the archive database is retrieved based on the archive features of the archives to be merged. The archive features of the archives to be merged are cover image characteristics or class centrality features, where the class centrality features are used to represent the average vector of feature vectors of multiple face images in the archives to be merged. If the archives to be merged use cover image features to represent their archive features, then the archive database is retrieved using the cover image features; if the archives to be merged use class centrality features to represent their archive features, then the archive database is retrieved using the class centrality features.

[0054] In specific embodiments, the database of files to be merged can be searched using mainstream vector retrieval methods, including but not limited to vector engine search (Milvus) and similar vector retrieval (Faiss).

[0055] In a specific embodiment, the archive database is searched based on the archives to be merged to obtain one or more pre-merge archives with different similarities. The pre-merge archives with similarity reaching a first preset threshold in the search results are then combined with the merged archives to form candidate archive pairs. Here, the first preset threshold is a relatively low threshold, the purpose of which is only to filter out archives with particularly low similarity. These archives often do not belong to the same target as the archives to be merged, and continuing to process them would only increase the amount of unnecessary computation.

[0056] In a specific embodiment, the search results of the database can be analyzed to determine whether the number of pre-mergeable files with a similarity level reaching a second preset threshold is greater than a third preset threshold. If so, only pre-mergeable files with a similarity level reaching the second preset threshold are selected to form candidate file pairs with the file to be merged. If not, only the pre-mergeable file with the highest similarity among the search results is selected to form a candidate file pair with the file to be merged. The second preset threshold here is a relatively high threshold, designed to determine if there are enough highly similar results. If so, only pre-mergeable files with high similarity are considered for forming file pairs; otherwise, only the pre-mergeable file with the highest similarity is considered. This optional strategy is a balancing strategy that helps reduce erroneous merging in subsequent steps.

[0057] S103: Obtain multi-dimensional feature information of the files to be merged and the files to be merged, as well as the interaction feature information of the candidate file pairs.

[0058] In a specific embodiment, the multi-dimensional feature information of the archives to be merged includes, but is not limited to: age, gender, whether a mask is worn, average quality score, cohesion, and number of images of the face images, wherein cohesion is used to characterize the average similarity of the feature vectors of any two face images in the archive.

[0059] In specific embodiments, the multi-dimensional feature information of the pre-merged archives includes, but is not limited to: the number of retrieved pre-merged archives and the number of pre-merged archives with a similarity greater than a fourth preset threshold. Here, the fourth threshold is higher than the first threshold, and this feature is used to characterize high-similarity retrieval results.

[0060] In specific embodiments, the interactive feature information of candidate file pairs includes, but is not limited to: the feature similarity between the two files in the candidate file pair, the cohesion after the two files are fused, the amount of reduction in cohesion, and the average quality score of the face image.

[0061] in:

[0062] Cohesion calculation: Assume the feature vector of the face images in the files to be merged is x = (x1, x2, ..., x...). n If x ∈ (x ...

[0063] Calculation of cohesion after merging two candidate archives:

[0064]

[0065] in,

[0066]

[0067] s a and s b The cohesion of the two files, n a and n b , where represents the number of face images in the two files, and sim represents the feature similarity between the two files.

[0068] Calculation of the reduction in cohesion after merging two candidate archives:

[0069] s ab -0.5×(s a +s b ).

[0070] Calculation of the average quality score of the face images after merging two candidate files: Assume the average quality score of the face images in the files to be merged is f. a The number of face images is n a The average quality of the facial images in the pre-merged archives is divided into f. b The number of face images is n b The average quality of the merged face images is then divided into...

[0071] S104: Input multi-dimensional feature information and interaction feature information into the trained machine learning model. The machine learning model outputs the confidence that two files in a candidate file pair belong to the same target.

[0072] In a specific embodiment, a pre-trained machine learning model is used. The model is input with multi-dimensional feature information of the files to be merged and the files to be merged, as well as the interaction feature information of the candidate file pairs. The machine learning model outputs the confidence score that two files in a candidate file pair belong to the same target. If the confidence score reaches a fifth preset threshold, the next step of file merging can be performed. The machine learning model can be an ensemble learning model, such as Random Forest, XGBoost, or LightGBM, to build a classification model.

[0073] S105: Determine whether candidate file pairs should be merged into a merged file based on the confidence level.

[0074] In a specific embodiment, since different candidate file pairs may contain the same pre-merge file (i.e., different files to be merged correspond to the same pre-merge file), a pruning strategy is also needed to achieve deduplication. The pruning strategy is as follows:

[0075] All candidate file pairs with a confidence level reaching the fifth preset threshold are selected, and the files to be merged and the files to be merged in all candidate file pairs are used to form a graph network; all files to be merged in the graph network are sorted from largest to smallest according to the number of face images to construct a pruning sequence; according to the pruning sequence, the files to be merged and all their associated files to be merged are merged into a merged file, and removed from the graph network after each merge.

[0076] Figure 2 A schematic diagram of a pruning strategy according to a specific embodiment of this application is shown, such as... Figure 2 As shown in the diagram, suppose there are three file pairs A, B, and C to be merged. File A forms candidate file pairs with pre-merged files A1, A2, and A3, respectively. File B forms candidate file pairs with pre-merged files A3, B1, B2, and B3, respectively. File C forms candidate file pairs with pre-merged files B3, C1, C2, and C3, respectively, as shown in the left figure. This forms the graph network structure shown in the right figure. Assuming the relationship between the number of face images in file pairs A, B, and C is A > B > C, a pruning sequence ABC is constructed. According to the pruning sequence, file pair A is first merged with pre-merged files A1, A2, and A3 to form merged file 1. Simultaneously, file pair A and pre-merged files A1, A2, and A3 are removed from the graph network. This process continues until all three file pairs A, B, and C are merged and deduplicated, resulting in merged file 1, merged file 2, and merged file 3.

[0077] Based on the pruning strategy described above, it can be seen that pre-merge file A3, associated with file pair B to be merged, is pruned, and pre-merge file B3, associated with file pair C to be merged, is pruned. This strategy is a conservative approach to prevent accidental merging of files from leading to continuous outward expansion and causing large errors.

[0078] It should be noted that in this embodiment, when there is a file to be merged, and all of its associated pre-merge files have been previously cut off, that is, when there are no pre-merge files to be merged, the file to be merged is not processed and is directly discarded.

[0079] It should be noted that the first preset threshold, the second preset threshold, the third preset threshold, the fourth preset threshold and the fifth preset threshold in this embodiment are not specifically limited, and can be set according to the actual file merging situation.

[0080] In summary, the file merging method of this application obtains multi-dimensional feature information of the files to be merged and the files to be merged, as well as the interaction feature information of candidate file pairs. Then, it uses a machine learning model to give the confidence level that two files belong to the same target. This results in higher accuracy and a higher merging rate, and can more effectively alleviate the phenomenon of multiple files with the same target in the archive. Furthermore, after selecting candidate file pairs with confidence levels that meet the preset threshold through the machine learning model, all candidate file pairs are constructed into a graph network, and pruning strategies are used to perform pruning operations, thereby achieving deduplication while merging files, resulting in a higher merging rate.

[0081] According to a second aspect of this application, based on the same concept, a file merging device is also proposed. Figure 3 A structural block diagram of a file merging apparatus according to an embodiment of this application is shown, such as... Figure 3 As shown, the device includes:

[0082] The acquisition module 201 is configured to acquire files to be merged, each file including a cover image and multiple face images;

[0083] The retrieval module 202 is configured to retrieve the archive database based on the archive to be merged, obtain the pre-merged archive containing multiple face images, and construct the archive to be merged and the pre-merged archive into a candidate archive pair;

[0084] Feature extraction module 203 is configured to obtain multi-dimensional feature information of the files to be merged and the files to be merged, as well as the interactive feature information of candidate file pairs;

[0085] The prediction module 204 is configured to input multi-dimensional feature information and interaction feature information into a trained machine learning model, and the machine learning model outputs the confidence that the two files in the candidate file pair belong to the same target.

[0086] Strategy module 205 is configured to determine whether candidate file pairs should be merged into a merged file based on confidence level.

[0087] According to a third aspect of this application, based on the same concept, a computer-readable storage medium is further proposed that stores a computer program which, when executed by a processor, implements the file merging method as described in the first aspect of this application.

[0088] In the embodiments of this application, it should be understood that the disclosed technical content can be implemented in other ways. The device / system / method embodiments described above are merely illustrative. For example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling or direct coupling or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0089] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0090] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0091] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0092] It is obvious that those skilled in the art can make various modifications and alterations to the embodiments of this application without departing from the spirit and scope of this application. In this way, this application also aims to cover such modifications and alterations if they fall within the scope of the claims and their equivalents. The word "comprising" does not exclude the presence of other elements or steps not listed in the claims. The simple fact that certain measures are described in mutually different dependent claims does not indicate that a combination of these measures cannot be used for profit. Any reference numerals in the claims should not be considered limiting in scope.

Claims

1. An archival merge method, characterized by, The method comprises the following steps: obtaining a to-be-merged archive, each of the to-be-merged archives comprising a cover image and a plurality of face images; retrieving an archive library according to the to-be-merged archives to obtain pre-merged archives containing a plurality of face images, and constructing the to-be-merged archives and the pre-merged archives into candidate archive pairs, comprising: retrieving the archive library according to the to-be-merged archives to obtain the pre-merged archives corresponding to one or more different similarities; constructing the pre-merged archives with a similarity reaching a first preset threshold in the retrieval results into the candidate archive pairs with the to-be-merged archives respectively; analyzing the retrieval results; determining whether the number of the pre-merged archives with a similarity reaching a second preset threshold is greater than a third preset threshold, if yes, only selecting the pre-merged archives with a similarity reaching the second preset threshold to construct the candidate archive pairs with the to-be-merged archives respectively, if no, only selecting the pre-merged archive with the largest similarity in the retrieval results to construct the candidate archive pairs with the to-be-merged archives; obtaining multi-dimensional feature information of the to-be-merged archives and the pre-merged archives and interaction feature information of the candidate archive pairs, wherein the multi-dimensional feature information of the to-be-merged archives and the pre-merged archives comprises age, gender, whether wearing a mask, average quality score, cohesiveness of face images in the to-be-merged archives, and the number of the pre-merged archives retrieved, the number of the pre-merged archives with a similarity greater than a fourth preset threshold; the interaction feature information of the candidate archive pairs comprises feature similarity between two archives in the candidate archive pairs and cohesiveness, reduction amount of cohesiveness and average quality score of face images after fusion of the two archives; inputting the multi-dimensional feature information and the interaction feature information into a trained machine learning model, and the machine learning model outputs a confidence degree that the two archives in the candidate archive pairs belong to the same target; determining whether the candidate archive pairs are merged into a merged archive according to the confidence degree.

2. The method of claim 1, wherein, The retrieval of the archive library according to the to-be-merged archives comprises: retrieving the archive library according to archive features of the to-be-merged archives, the archive features being cover image characteristics or class center features, wherein the class center features are used to represent the average vector of the feature vectors of the plurality of face images in the to-be-merged archives.

3. The method of claim 1, wherein, The expression of the cohesiveness of the two archives in the candidate archive pairs after fusion is specifically: ； wherein, ； is the number of face images in the first archive, and is the cohesion of the first archive, and is the number of face images in the second archive, is the feature similarity of the two archives. The expression of the reduction amount of the cohesiveness of the two archives in the candidate archive pairs after fusion is specifically: 。 4. The method of claim 1, wherein, The determination of whether the candidate archive pairs are merged into a merged archive according to the confidence degree comprises: screening all the candidate archive pairs with a confidence degree reaching a fifth preset threshold, and constructing the to-be-merged archives and the pre-merged archives in the all candidate archive pairs into a graph network; sorting all the to-be-merged archives in the graph network according to the number of face images from large to small to construct a pruning sequence; According to the pruning sequence, the to-be-merged archives and all the pre-merged archives associated therewith are merged into the merged archive one by one, and are removed from the graph network after each merging.

5. An archival merge apparatus, comprising: Comprise: An acquisition module configured to acquire to-be-merged archives, each of the to-be-merged archives comprising a cover image and multiple face images; A retrieval module configured to retrieve an archive library according to the to-be-merged archives, obtain pre-merged archives containing multiple face images, and construct candidate archive pairs from the to-be-merged archives and the pre-merged archives, comprising: Retrieving the archive library according to the to-be-merged archives to obtain the pre-merged archives corresponding to one or more different similarities; The pre-merged archives in the retrieval results with a similarity reaching a first preset threshold are respectively constructed into the candidate archive pairs with the to-be-merged archives; Analyzing the retrieval results; Determine whether the number of pre-merged archives with a similarity reaching a second preset threshold is greater than a third preset threshold, if yes, only select the pre-merged archives with a similarity reaching the second preset threshold to construct into the candidate archive pairs with the to-be-merged archives; if not, only select the pre-merged archive with the largest similarity in the retrieval results to construct into the candidate archive pairs with the to-be-merged archives; A feature extraction module configured to obtain multi-dimensional feature information of the to-be-merged archives and the pre-merged archives and interaction feature information of the candidate archive pairs, wherein the multi-dimensional feature information of the to-be-merged archives and the pre-merged archives comprises: the age, gender, whether wearing a mask, average quality score, cohesiveness, and number of images of the face images in the to-be-merged archives, wherein the cohesiveness is used to represent the average value of the similarity of the feature vectors of any two face images in the archive; and the number of retrieved pre-merged archives, the number of pre-merged archives with a similarity greater than a fourth preset threshold; The interaction feature information of the candidate archive pairs comprises: the feature similarity between the two archives in the candidate archive pair, and the cohesiveness, reduction amount of cohesiveness, and average quality score of face images after fusion of the two archives; A prediction module configured to input the multi-dimensional feature information and the interaction feature information into a trained machine learning model, wherein the machine learning model outputs a confidence degree of the two archives in the candidate archive pair belonging to the same target; A strategy module configured to determine whether the candidate archive pair is merged into a merged archive according to the confidence degree.

6. A computer-readable storage medium storing a computer program, wherein the computer program is executed by a processor to implement the method of any one of claims 1-4.

Images