Portrait archive processing method and apparatus, storage medium, and electronic device

Abstract

This application discloses a method, apparatus, storage medium, and electronic device for processing portrait archives. The method includes: performing temporary spatiotemporal extraction on a set of capture data of a target portrait archive within a preset time period from a set of portrait archives in a preset area to obtain a set of temporary spatiotemporal data of the target portrait archive; determining a target temporary spatiotemporal data with missing capture points based on the correlation between capture points in the set of temporary spatiotemporal data, wherein the missing capture points are predicted capture points with missing capture data in the target temporary spatiotemporal data; recalling unfiled capture images that match the missing capture points to the target portrait archive to update the target portrait archive, wherein the unfiled capture images are not capture images of any portrait archive in the set of portrait archives; and performing file merging processing on the target portrait archive and the candidate portrait archive when there is a candidate portrait archive that matches the target temporary spatiotemporal data in the set of portrait archives, and it is determined that file merging is allowed based on the similarity between the target portrait archive and the candidate portrait archive.

Description

Technical Field

[0001] This application relates to the field of image processing, and more specifically, to a method and apparatus for processing portrait archives, a storage medium, and an electronic device. Background Technology

[0002] Facial recognition technology creates portrait profiles of different individuals, allowing for the reconstruction of their trajectories. This reconstruction can then be used to locate key information such as fellow travelers. In existing technologies, portrait images captured by cameras or other image acquisition devices are first processed using deep learning to generate feature vectors. Similarity calculations are then performed based on these feature vectors to create portrait profiles.

[0003] However, due to differences in capture distance, capture angle, and capture equipment, the similarity of different portrait images of the same subject may not meet the criteria for clustering, resulting in multiple files for one person and making it impossible to completely reconstruct the portrait trajectory. Therefore, it can be seen that the portrait file processing methods in related technologies suffer from low recall rates due to the low accuracy of portrait image recognition for the same subject. Summary of the Invention

[0004] This application provides a method and apparatus for processing portrait archives, a storage medium, and an electronic device to at least solve the problem of low recall rate of portrait archives due to low accuracy of portrait image recognition of the same object in related art processing methods.

[0005] According to one aspect of the embodiments of this application, a method for processing portrait archives is provided, comprising: performing temporary spatiotemporal extraction on a set of capture data of a target portrait archive within a set of portrait archives corresponding to a preset area within a preset time period to obtain a set of temporary spatiotemporal spaces within the target portrait archive, wherein a temporary spatiotemporal space contains a segment of continuous capture data associated with corresponding spatiotemporal information, and the spatiotemporal information of a capture data segment is used to record the capture checkpoint and capture time corresponding to the capture data segment; determining a target temporary spatiotemporal space with a missed capture checkpoint in the set of temporary spatiotemporal spaces based on the association relationship between different capture checkpoints under the set of temporary spatiotemporal spaces, wherein the target temporary spatiotemporal space under the The missing capture point is the predicted capture point in the target's temporary spatiotemporal context that lacks corresponding capture data. Unfiled capture images matching the missing capture points in the target's temporary spatiotemporal context are recalled to the target portrait file to update the target portrait file. The unfiled capture images do not belong to any portrait file in the set of portrait files. If a candidate portrait file matching the target's temporary spatiotemporal context exists in the set of portrait files, and the similarity between the target portrait file and the candidate portrait file determines that merging the target portrait file and the candidate portrait file is permissible, then the target portrait file and the candidate portrait file are merged.

[0006] According to another aspect of the embodiments of this application, a processing apparatus for portrait archives is also provided, comprising: an extraction unit, configured to perform temporary spatiotemporal extraction on a set of capture data of a target portrait archive within a set of portrait archives corresponding to a preset area within a preset time period, to obtain a set of temporary spatiotemporal spaces within the target portrait archive, wherein a temporary spatiotemporal space contains a segment of continuous capture data associated with corresponding spatiotemporal information, and the spatiotemporal information of a capture data segment is used to record the capture checkpoint and capture time corresponding to the capture data segment; and a first determining unit, configured to determine a target temporary spatiotemporal space in the set of temporary spatiotemporal spaces that has a missing capture checkpoint based on the association relationship between different capture checkpoints under the set of temporary spatiotemporal spaces, wherein the target temporary spatiotemporal space The missing capture points are the predicted capture points in the target temporary spatiotemporal context that lack corresponding capture data; the recall unit is used to recall the unfiled capture images that match the missing capture points in the target temporary spatiotemporal context to the target portrait file to update the target portrait file, wherein the unfiled capture images do not belong to any portrait file in the set of portrait files; the merging unit is used to merge the target portrait file and the candidate portrait file when there is a candidate portrait file that matches the target temporary spatiotemporal context in the set of portrait files, and the similarity between the target portrait file and the candidate portrait file determines that the target portrait file and the candidate portrait file can be merged.

[0007] In an exemplary embodiment, the extraction unit includes: a sorting module, configured to sort the set of captured data according to the corresponding capture time to obtain a target capture trajectory corresponding to the target portrait file; and a first extraction module, configured to perform temporary spatiotemporal extraction on the target capture trajectory to obtain the set of temporary spatiotemporal data.

[0008] In an exemplary embodiment, the extraction unit includes: a second extraction module, configured to extract each segment of continuous capture data that meets preset extraction conditions from the set of capture data to obtain the set of temporary spatiotemporal data; wherein, the preset extraction conditions include: the distance between capture checkpoints corresponding to any two adjacent capture data is less than or equal to a first distance, or there is an intersection between preset information surfaces of capture checkpoints corresponding to any two adjacent capture data, or there is an intersection between keywords contained in the checkpoint identifiers of capture checkpoints corresponding to any two adjacent capture data, wherein the preset information surface is the information surface corresponding to the second distance; the time difference between capture times corresponding to any two adjacent capture data is less than or equal to a first time threshold, and the total capture duration is greater than or equal to a second time threshold; and it does not belong to any temporary spatiotemporal space within the target portrait file.

[0009] In an exemplary embodiment, the second extraction module includes: a sorting submodule, configured to sort the set of capture data according to the corresponding capture time to obtain a target capture trajectory corresponding to the target portrait file; a first extraction submodule, configured to extract keywords from the checkpoint identifier of the capture checkpoint corresponding to each capture data in the set of capture data to obtain keywords contained in the checkpoint identifier of the capture checkpoint corresponding to each capture data; an updating submodule, configured to update each capture data in the target capture trajectory using the keywords contained in the checkpoint identifier of the capture checkpoint corresponding to each capture data to obtain an updated capture trajectory, wherein each capture data in the updated capture trajectory is used to record the capture checkpoint corresponding to each capture data, the keywords contained in the checkpoint identifier of the corresponding capture checkpoint, and the capture time; and a second extraction submodule, configured to perform temporary spatiotemporal extraction on the updated capture trajectory to obtain the set of temporary spatiotemporal data.

[0010] In an exemplary embodiment, the first determining unit includes: a first determining module, configured to determine the total number of times each capture checkpoint in a set of capture checkpoints appears within the set of temporary spatiotemporal spaces and the co-occurrence count of each capture checkpoint pair in the set of capture checkpoints, wherein the set of capture checkpoints comprises all capture checkpoints under the set of temporary spatiotemporal spaces, and the co-occurrence count of each capture checkpoint pair is the total number of times the two capture checkpoints in each capture checkpoint pair appear in the same temporary spatiotemporal space; and a second determining module, configured to determine, based on the total number of times each capture checkpoint appears within the set of temporary spatiotemporal spaces and the co-occurrence count of each capture checkpoint pair, a number related to the capture checkpoints in the set of temporary spatiotemporal spaces. The association probability corresponding to each capture checkpoint, wherein the association probability corresponding to each capture checkpoint includes the association probability between each capture checkpoint and other capture checkpoints in the group of capture checkpoints besides each capture checkpoint, and the association probability between each capture checkpoint and the other capture checkpoints is the probability of the other capture checkpoints appearing when each capture checkpoint appears in a temporary spacetime within the target portrait file; the third determining module is used to determine the target temporary spacetime with a missed checkpoint in the group of temporary spacetime based on the association probability corresponding to the capture checkpoints in each temporary spacetime of the group of temporary spacetimes.

[0011] In an exemplary embodiment, the third determining module includes: an execution submodule, configured to perform the following operations on each temporary spacetime as a current temporary spacetime to obtain the target temporary spacetime: determining a set of current capture checkpoints appearing in the current temporary spacetime and a set of candidate capture checkpoints not appearing in the current temporary spacetime; determining the missed pass probability corresponding to each candidate capture checkpoint based on the association probability between each current capture checkpoint in the set of current capture checkpoints and each candidate capture checkpoint in the set of candidate capture checkpoints; and determining the current temporary spacetime as the target temporary spacetime if there is a missed pass checkpoint in the set of candidate capture checkpoints with a missed pass probability greater than or equal to a preset probability threshold.

[0012] In one exemplary embodiment, the apparatus further includes: a first filtering unit, configured to, before recalling the unfiled captured images that match the missed checkpoints under the target temporary spatiotemporal space to the target portrait file, filter out unfiled captured images whose capture time is within the time range corresponding to the target temporary spatiotemporal space from the set of unfiled captured images, to obtain a set of candidate captured images; and a second determining unit, configured to, if there is a target captured image in the set of candidate captured images with a similarity greater than or equal to a first similarity threshold with the target portrait file, determine the target captured image as an unfiled captured image that matches the missed checkpoints under the target temporary spatiotemporal space, wherein the first similarity threshold is negatively correlated with the missed probability corresponding to the missed checkpoints under the target temporary spatiotemporal space.

[0013] In one exemplary embodiment, the apparatus further includes: a second filtering unit, configured to, after recalling the unfiled captured images that match the missed checkpoints under the target temporary spatiotemporal space to the target portrait archive, filter from the set of portrait archives portrait archives portrait archives containing portrait archives whose corresponding capture time is within the time range corresponding to the target temporary spatiotemporal space and whose corresponding capture checkpoint matches the missed checkpoints under the target temporary spatiotemporal space, to obtain the candidate portrait archive; and a third determining unit, configured to, if the similarity between the target portrait archive and the candidate portrait archive is greater than or equal to a second similarity threshold, determine that the target portrait archive and the candidate portrait archive are allowed to be merged, wherein the second similarity threshold is negatively correlated with the missed probability corresponding to the missed checkpoints under the target temporary spatiotemporal space.

[0014] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, wherein a computer program is stored in the computer-readable storage medium, and the computer program is configured to execute the above-described method for processing portrait files when it is run.

[0015] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the above-mentioned method for processing portrait files through the computer program.

[0016] In this embodiment, a method is adopted to mine the potential missing capture points within the temporary spatiotemporal space of the portrait archive. Temporary spatiotemporal extraction is performed on a set of capture data from a target portrait archive within a set of portrait archives corresponding to a preset area within a preset time period. This yields a set of temporary spatiotemporal spaces within the target portrait archive. Each temporary spatiotemporal space contains a segment of continuous capture data associated with corresponding spatiotemporal information. The spatiotemporal information of a single capture data segment records the capture point and capture time corresponding to that segment. Based on the association between different capture points within a set of temporary spatiotemporal spaces, a target temporary spatiotemporal space containing missing capture points is determined. The missing capture points within the target temporary spatiotemporal space are predicted capture points that lack corresponding capture data within the target temporary spatiotemporal space. Unfiled capture images matching the missing capture points within the target temporary spatiotemporal space are recalled to the target portrait archive to update it. These unfiled capture images do not belong to the target portrait archive. A capture image of any portrait file in a set of portrait files; if a candidate portrait file exists in the set of portrait files that matches the target's temporary spatiotemporal location, and the target and candidate portrait files are allowed to be merged based on their similarity, the target and candidate portrait files are merged. Since there is a certain correlation between the capture points of different temporary spatiotemporal locations in a portrait file, the missing capture points in each temporary spatiotemporal location of a portrait file can be identified based on the discovered correlation. Thus, the capture data corresponding to the missing capture points can be found, and corresponding discarded images can be recalled and the files can be merged. This can improve the accuracy of portrait image recognition of the same object, achieve the technical effect of improving the recall rate of portrait file merging, and solve the problem of low recall rate of portrait file merging due to low accuracy of portrait image recognition of the same object in related technologies. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the hardware environment for an optional method of processing portrait files according to an embodiment of this application;

[0020] Figure 2 This is a flowchart illustrating an optional method for processing portrait files according to an embodiment of this application;

[0021] Figure 3 This is a schematic diagram of an optional method for processing portrait files according to an embodiment of this application;

[0022] Figure 4 This is a schematic diagram of another optional method for processing portrait files according to an embodiment of this application;

[0023] Figure 5 This is a flowchart illustrating another optional method for processing portrait files according to an embodiment of this application;

[0024] Figure 6 This is a structural block diagram of an optional human image file processing device according to an embodiment of this application;

[0025] Figure 7 This is a structural block diagram of an optional electronic device according to an embodiment of this application. Detailed Implementation

[0026] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present application.

[0027] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0028] According to one aspect of the embodiments of this application, a method for processing portrait archives is provided. Optionally, in this embodiment, the above-described method for processing portrait archives can be applied to, for example... Figure 1 The hardware environment shown includes the camera device 102 and the server 104. For example... Figure 1As shown, server 104 is connected to shooting device 102 via a network and can be used to provide services (such as application services) to shooting device 102 or clients installed on shooting device. A database can be set up on the server or independently of the server to provide data storage services for server 104.

[0029] The aforementioned network may include, but is not limited to, at least one of the following: wired network, wireless network. The aforementioned wired network may include, but is not limited to, at least one of the following: wide area network, metropolitan area network, local area network. The aforementioned wireless network may include, but is not limited to, at least one of the following: Wi-Fi (Wireless Fidelity), Bluetooth. The shooting device 102 is not limited to a camera, etc.

[0030] The method for processing portrait files in this embodiment can be executed by server 104. Figure 2 This is a flowchart illustrating an optional method for processing portrait archives according to an embodiment of this application, as shown below. Figure 2 As shown, the process of this method may include the following steps:

[0031] Step S202: Temporary spatiotemporal extraction is performed on a set of capture data of a target portrait file in a set of portrait files corresponding to a preset area within a preset time period to obtain a set of temporary spatiotemporal data within the target portrait file. Among them, a temporary spatiotemporal data contains a segment of continuous capture data associated with corresponding spatiotemporal information. The spatiotemporal information of a capture data is used to record the capture checkpoint and capture time corresponding to a capture data.

[0032] The image archive processing method in this embodiment can be applied to scenarios involving image clustering of images captured by a camera. Here, an image archive refers to an archive formed after image clustering of captured images based on facial recognition technology. Image clustering technology can be used to cluster captured images of the same person based on facial recognition technology to form an image archive, and then reconstruct the movement trajectory of the corresponding person based on the image archive.

[0033] Current facial image clustering technology relies on capturing images of people from cameras within cities. This data is then analyzed using deep learning to generate feature vectors, which are used to calculate similarity and create image clusters. However, due to variations in capture distance, angle, and equipment, the similarity of images of the same person may not meet the clustering criteria, resulting in multiple clusters for the same person and making it impossible to fully reconstruct the person's trajectory.

[0034] For example, existing human behavior modeling and recognition methods based on prior knowledge clustering possess self-learning capabilities, continuously improving the classifier as observed samples are classified. If outliers exist in the cluster distribution obtained during the human behavior modeling sub-process, they are treated as abnormal behavior models, and warnings can be issued as needed during system design. While these methods consider the significant impact of prior knowledge on human behavior recognition results and employ improved clustering methods for modeling and recognizing human behavior, laying the foundation for abnormal behavior identification, they require prior knowledge of each individual's behavior. Recognizing behavior based on known prior knowledge is problematic; if no prior knowledge exists, it is considered abnormal. In reality, it's impossible to obtain all behavioral characteristics of every individual, and the varying camera angles can easily introduce significant errors in human behavior recognition, making this method difficult to apply for human clustering.

[0035] Furthermore, existing human trajectory processing methods involve acquiring face trajectories, a first human trajectory, a second human trajectory, and a third human trajectory; binding the face trajectory and the first human trajectory to obtain a first trajectory mapping relationship; retrieving the face trajectory from a face database to obtain the target permanent identity corresponding to the face trajectory; clustering the first and second human trajectories and determining the temporary identity corresponding to the clustering results to obtain a second trajectory mapping relationship; retrieving the third human trajectory from the first and second human trajectories to obtain a third trajectory mapping relationship; and binding the target permanent identity to the corresponding human trajectory based on the first, second, and third trajectory mapping relationships and the target permanent identity. However, this method merely clusters multiple sets of human data and the data resulting from the association of face and human bodies, reducing the amount of clustered data but failing to effectively improve the accuracy and recall of the clustering.

[0036] To at least partially solve the aforementioned technical problems, in this embodiment, after determining the portrait archives using portrait clustering technology, the association between capture points in the portrait archives can be used to identify potential missed capture points in each temporary spatiotemporal space. Then, discarded images and archives that meet the similarity threshold within the time range of the temporary spatiotemporal space corresponding to the missed capture points are searched for, and these discarded images are recalled and merged, thereby achieving the recall of portrait archives. Here, the temporary spatiotemporal space can be determined based on the trajectory of the corresponding portrait archives. Here, discarded images refer to images that are archived separately due to capture angle, capture clarity, or other reasons.

[0037] In this embodiment, a set of captured data within a preset time period from a set of target portrait files corresponding to a preset area can be temporarily extracted to obtain a set of temporary spatiotemporal data within the target portrait file. Here, a temporary spatiotemporal space can contain a segment of continuous captured data associated with corresponding spatiotemporal information. The captured data can be captured portrait images, and a set of captured data can contain multiple captured data. The spatiotemporal information of a single captured data can be used to record the capture checkpoint and capture time corresponding to that single captured data.

[0038] Optionally, the aforementioned target portrait profile can be a single portrait profile formed by calculating the similarity of portrait images within a preset region and time period using methods such as deep learning. A set of portrait profiles can contain multiple portrait profiles. It should be noted that, in this embodiment, all operations performed on the target portrait profile can also be performed on other portrait profiles within the same set of portrait profiles.

[0039] Step S204: Based on the correlation between different capture checkpoints under a set of temporary time and space, determine a target temporary time and space with missing checkpoints in a set of temporary time and space. The missing checkpoints under the target temporary time and space are the predicted checkpoints that are missing corresponding capture data under the target temporary time and space.

[0040] Because a person's trajectory is unpredictable, within a preset time period, a person may pass through the same checkpoint multiple times and be captured by the corresponding camera. Therefore, each temporary temporal space within a set of temporary temporal spaces can contain at least one capture checkpoint, and different temporary temporal spaces within a set of temporary temporal spaces can contain the same capture checkpoints. There can be certain correlations between the capture checkpoints within the temporary temporal spaces of the target person's image file.

[0041] Optionally, the aforementioned correlation can be determined based on the total number of times each capture point appears in the target image file and the number of times it appears simultaneously with other capture points.

[0042] In this embodiment, based on the correlation between different capture points under a set of temporary temporal spaces, a target temporary temporal space with missing capture points can be determined. Here, the missing capture points under the target temporary temporal space can be the predicted capture points that lack corresponding capture data under the target temporary temporal space.

[0043] Optionally, based on the correlation between different capture points in the temporary space-time within the target portrait file, and combined with the capture points appearing in each temporary space-time within the target portrait file, it is possible to predict the capture points with a high probability of appearing in each temporary space-time within the target portrait file, and determine the capture points that are missing corresponding capture data, i.e., the missing capture points, based on the capture data corresponding to that temporary space-time.

[0044] Step S206: Recall the unfiled captured images that match the missing checkpoints in the target temporary space-time to the target portrait file to update the target portrait file. The unfiled captured images do not belong to any portrait file in a set of portrait files.

[0045] Considering the possibility that each captured image corresponding to a missed checkpoint may not be aggregated into a portrait file, after identifying the missed checkpoints in the target temporary time-space, the unfiled captured images matching the target temporary time-space can be determined based on the information of the target temporary time-space (e.g., time range). Here, unfiled captured images may not belong to any portrait file in a set of portrait files. Under normal circumstances, unfiled captured images are generally considered as unusable data.

[0046] In this embodiment, unfiled captured images that match the target's temporary spatiotemporal context can be recalled to update a set of portrait profiles.

[0047] Optionally, to improve the accuracy of image retrieval, before performing retrieval processing, the similarity between the identified unfiled snapshots that temporarily match the target and the snapshots in the target's portrait file can be calculated, and the unfiled snapshots with similarity values ​​that meet the conditions can be retrieved.

[0048] Step S208: If a candidate image file that temporarily matches the target image file exists in a set of image files, and the target image file and the candidate image file are allowed to be merged based on the similarity between the target image file and the candidate image file, then the target image file and the candidate image file are merged.

[0049] When several captured images corresponding to a missed checkpoint are grouped into a single portrait profile, after image retrieval and updating the portrait profile, candidate portrait profiles matching the target temporary spatiotemporal space can be determined based on the target temporary spatiotemporal space information (e.g., time range). Here, the time range of the candidate portrait profile's temporary spatiotemporal space can be the same as or similar to the target temporary spatiotemporal space's time range, and the captured data of the temporary spatiotemporal spaces with the same or similar time ranges to the target temporary spatiotemporal space correspond to the missed checkpoint.

[0050] Optionally, to improve the accuracy of file merging, the similarity between the target person's image file and the identified candidate image files can be calculated, and files with similarity values ​​that meet the criteria can be merged. The similarity calculation can be performed on the images in the files.

[0051] In this embodiment, if there is a candidate image file that temporarily matches the target in a set of image files, and the target image file and the candidate image file are allowed to be merged based on the similarity between the target image file and the candidate image file, then the target image file and the candidate image file can be merged.

[0052] Through the above steps, temporary spatiotemporal extraction is performed on the capture data of the target portrait file within a set of portrait files corresponding to the preset area within a preset time period, resulting in a set of temporary spatiotemporal spaces within the target portrait file. Each temporary spatiotemporal space contains a segment of continuous capture data associated with corresponding spatiotemporal information. The spatiotemporal information of a single capture data segment records the capture checkpoint and capture time corresponding to that segment. Based on the correlation between different capture checkpoints within a set of temporary spatiotemporal spaces, a target temporary spatiotemporal space with missing capture checkpoints is identified. The missing capture checkpoints within the target temporary spatiotemporal space are predicted checkpoints that lack corresponding capture data within the target temporary spatiotemporal space. The target temporary spatiotemporal space is then linked to the target temporary spatiotemporal space. Unfiled images captured during spatiotemporal matching at checkpoints are recalled to the target portrait archive to update it. These unfiled images do not belong to any portrait archive within a set of portrait archives. If a candidate portrait archive exists within a set of portrait archives that temporarily matches the target in spatiotemporal matching, and the similarity between the target and candidate portrait archives determines that merging the target and candidate portrait archives is permissible, then the target and candidate portrait archives are merged. This addresses the problem of low recall rates in portrait archive processing methods in related technologies due to low accuracy in recognizing images of the same object, thus improving the recall rate of portrait archive merging.

[0053] In an exemplary embodiment, a set of captured data within a preset time period from a set of portrait files corresponding to a preset area is temporarily spatiotemporally extracted to obtain a set of temporary spatiotemporal data within the target portrait file, including:

[0054] S11, Sort a set of captured data according to the corresponding capture time to obtain the target capture trajectory corresponding to the target portrait file;

[0055] S12, perform temporary spatiotemporal extraction on the target capture trajectory to obtain a set of temporary spatiotemporal data.

[0056] When performing temporary spatiotemporal extraction of a set of captured data from a target person's image file within a preset time period, the extraction can be based on the capture trajectory corresponding to the target image file. Considering that a person's movement trajectory is time-dependent, the target capture trajectory can be obtained by sorting a set of captured data from the target person's image file within the preset time period according to the corresponding capture time. The target capture trajectory can include the capture point and the capture time.

[0057] For example, taking the target portrait file A as an example, the capture data of file A in the past N days is sorted according to the capture time, and the capture trajectory of file A in the past N days is as follows:

[0058]

[0059] Among them, C A1 For the first snapshot data of file A within the last N days, T is the capture point corresponding to the capture camera. A1 The corresponding capture time is given, and file A has been captured n times in the past N days.

[0060] Optionally, the aforementioned capture trajectory can be determined based on the continuity of the capture time and capture location of a set of capture data. Within a preset time period, the target portrait file can have multiple capture trajectories.

[0061] After determining the target's capture trajectory, a temporary spatiotemporal extraction can be performed on the target's capture trajectory to obtain a set of temporary spatiotemporal segments within the target's image file. Here, each temporary spatiotemporal segment can contain a capture trajectory segment.

[0062] This embodiment determines the capture trajectory by the sequence of capture times, thereby extracting the temporary spatiotemporal space of the portrait file, which can improve the accuracy of extracting the temporary spatiotemporal space and thus improve the processing efficiency of the portrait file.

[0063] In an exemplary embodiment, a set of captured data within a preset time period from a set of portrait files corresponding to a preset area is temporarily spatiotemporally extracted to obtain a set of temporary spatiotemporal data within the target portrait file, including:

[0064] S21, extract each segment of continuous capture data that meets the preset extraction conditions from a set of capture data to obtain a set of temporary spatiotemporal data;

[0065] The preset extraction conditions include: the distance between any two adjacent capture data points corresponding to capture checkpoints is less than or equal to a first distance; or there is an intersection between the preset information surfaces of any two adjacent capture data points corresponding to capture checkpoints; or there is an intersection between the keywords contained in the checkpoint identifiers of any two adjacent capture data points corresponding to capture checkpoints. The preset information surface is the information surface corresponding to the second distance; the time difference between any two adjacent capture data points corresponding to capture times is less than or equal to a first time threshold, and the total capture duration is greater than or equal to a second time threshold; and it does not belong to any temporary spacetime within the target portrait file.

[0066] To improve the accuracy of temporary spatiotemporal data, a preset extraction condition can be set to extract temporary spatiotemporal data from continuous snapshots. After determining a set of snapshots for the target portrait file, each segment of continuous snapshots that meets the preset extraction condition can be extracted from the set of snapshots to obtain a set of temporary spatiotemporal data. Here, the preset extraction condition may include: the distance between any two adjacent snapshots corresponding to the snapshot checkpoints is less than or equal to a first distance; or there is an intersection between the preset information surfaces of any two adjacent snapshots corresponding to the snapshot checkpoints; or there is an intersection between the keywords contained in the checkpoint identifiers of any two adjacent snapshots corresponding to the snapshots. The preset information surface may be the information surface corresponding to a second distance.

[0067] Optionally, the preset extraction conditions, in addition to the conditions mentioned above, may also include that the time difference between the capture times corresponding to any two adjacent capture data is less than or equal to a first time threshold, and the total capture duration is greater than or equal to a second time threshold. To avoid overlap of capture data from multiple temporary spatiotemporal locations within the target portrait file, the preset extraction conditions, in addition to the conditions mentioned above, may also include that the capture data does not belong to any temporary spatiotemporal location within the target portrait file.

[0068] For example, taking a target portrait file as file A, if file A contains a segment of continuous capture data that meets the following three conditions: the straight-line distance between any two capture points is less than m1 meters (i.e., the first distance) or the distance between any two capture points is m2 meters (i.e., the second distance); there is an intersection of AOI (Area of ​​Interest) information or an intersection of keyword information between any two capture points; the capture intervals before and after the continuous capture data are all less than t1 (i.e., the first time threshold), and the total capture duration is greater than t2 (i.e., the second time threshold); this temporary spatiotemporal space is not contained by other temporary spatiotemporal spaces of the file, then the spatiotemporal information of this segment of continuous capture data constitutes a temporary spatiotemporal space of file A. Assume the continuous capture data {C} A2 ,T A2},{C A3 ,T A3},{C A4 ,T A4 Let {C} be a temporary spacetime for file A. A2 ,T A2},{C A3 ,T A3},{C A4 ,T A4 This continuous capture data must satisfy checkpoint C. A2 C A3 C A4 The straight-line distance between any two checkpoints is less than m1 meters or checkpoint C A2 CA3 , C A4 There is an intersection in the m2-meter AOI information between two pairs of checkpoints or checkpoint C A2 , C A3 , C A4 There is an intersection in the checkpoint keyword information between two pairs of checkpoints and it satisfies T A3 -T A2 <t1, T A4 -T A3 <t1 and T A4 -T A1 >t2, and at the same time, the continuous capture data {C A2 , T A2},{C A3 , T A3},{C A4 , T A4} is not included in other temporary time-spaces of file A.

[0069] According to the above method, there can be three temporary time-spaces in file A in the recent N days as follows:

[0070]

[0071] Through this embodiment, by setting preset extraction conditions to determine the division rules of temporary time-spaces, the accuracy of dividing temporary time-spaces can be improved.

[0072] In an exemplary embodiment, each piece of continuous capture data that meets the preset extraction conditions is extracted from a set of capture data to obtain a set of temporary time-spaces, including:

[0073] S31, sort a set of capture data according to the corresponding capture time to obtain a target capture trajectory corresponding to the target portrait file;

[0074] S32, perform keyword extraction on the checkpoint identifier of the checkpoint corresponding to each capture data in a set of capture data to obtain the keywords included in the checkpoint identifier of the checkpoint corresponding to each capture data;

[0075] S33, update each capture data in the target capture trajectory with the keywords included in the checkpoint identifier of the checkpoint corresponding to each capture data to obtain an updated capture trajectory, where each capture data in the updated capture trajectory is used to record the checkpoint corresponding to each capture data, the keywords included in the checkpoint identifier of the corresponding checkpoint, and the capture time;

[0076] S34, perform temporary time-space extraction on the updated capture trajectory to obtain a set of temporary time-spaces.

[0077] When extracting keywords from the checkpoint identifier corresponding to each capture data point, the capture data can first be sorted according to the corresponding capture time to obtain the target capture trajectory. Then, keywords can be extracted from the checkpoint identifier corresponding to each capture data point in the set of capture data to obtain the keywords contained in the checkpoint identifier corresponding to each capture data point. Here, the extraction method and content of the target capture trajectory can be the same as in the previous embodiment, and will not be repeated here. The checkpoint identifier can be the name of the checkpoint, such as the Chinese name. A checkpoint can contain multiple keywords.

[0078] Optionally, the above-mentioned method of extracting keywords can be to first segment the checkpoint identifier into words, and then extract keywords from the co-occurrence information (semantics) among multiple word segments, or it can be to directly extract keywords from the checkpoint identifier.

[0079] For example, taking a checkpoint named "Hangzhou Civic Center North Gate Facing East" as an example, we can use word segmentation methods such as jieba (a third-party Chinese word segmentation function library) to obtain the three words "Hangzhou Civic Center", "North Gate", and "Facing East". Then, we can use keyword extraction algorithms such as TextRank (a graph-based ranking algorithm for keyword extraction and document summarization) to extract the keyword "Hangzhou Civic Center", and finally obtain the keyword for each checkpoint.

[0080] In this embodiment, the keywords contained in the checkpoint identifier corresponding to each capture data point can be used to update each capture data point in the target capture trajectory, resulting in an updated capture trajectory. Then, temporary spatiotemporal extraction is performed on the updated capture trajectory to obtain a set of temporary spatiotemporal data. Here, each capture data point in the updated capture trajectory can be used to record the corresponding capture checkpoint, the keywords contained in the checkpoint identifier, and the capture time.

[0081] For example, taking the target portrait file A as an example, when the target capture trajectory of file A is as follows:

[0082]

[0083] After extracting the checkpoint identifier for each checkpoint and extracting keywords from the checkpoint identifier, the updated file trajectory information can be obtained as follows:

[0084]

[0085] Where K An It is the C-type gate. An The extracted keywords.

[0086] In this embodiment, the keywords extracted from each checkpoint are added to the corresponding capture trajectory, which can improve the convenience of dividing temporary space and time.

[0087] In an exemplary embodiment, based on the correlation between different capture checkpoints under a set of temporary temporal spaces, a target temporary temporal space with missed capture checkpoints is determined, including:

[0088] S41, determine the total number of times each capture checkpoint in a set of capture checkpoints appears in a set of temporary spacetime and the co-occurrence number of each capture checkpoint pair in a set of capture checkpoints, wherein a set of capture checkpoints is all capture checkpoints under a set of temporary spacetime, and the co-occurrence number of each capture checkpoint pair is the total number of times the two capture checkpoints in each capture checkpoint pair appear in the same temporary spacetime.

[0089] S42, based on the total number of times each capture checkpoint appears in a set of temporary spatiotemporal spaces and the number of times each capture checkpoint pair co-occurs, determine the association probability corresponding to each capture checkpoint. The association probability corresponding to each capture checkpoint includes the association probability between each capture checkpoint and other capture checkpoints in the set of capture checkpoints besides each capture checkpoint. The association probability between each capture checkpoint and other capture checkpoints is the probability of other capture checkpoints appearing when each capture checkpoint appears in a temporary spatiotemporal space within the target portrait file.

[0090] S43, based on the association probability corresponding to the capture checkpoint under each temporary spacetime in a set of temporary spacestimes, determine the target temporary spacetime with a missed checkpoint in a set of temporary spacestimes.

[0091] In this embodiment, after determining all temporary spatiotemporal locations of the target portrait file, the total number of times each capture point in a set of capture points appears within that set of temporary spatiotemporal locations, and the co-occurrence count of each capture point pair within that set of capture points, can be determined. Here, a set of capture points can be all capture points within a set of temporary spatiotemporal locations. The co-occurrence count of each capture point pair can be the total number of times the two capture points in each capture point pair appear in the same temporary spatiotemporal location.

[0092] Optionally, a connectivity graph can be generated based on the total number of times each capture checkpoint in a set of capture checkpoints appears within a set of temporary spatiotemporal spaces and the co-occurrence count of each pair of capture checkpoints in the set of capture checkpoints. Here, the connectivity graph includes nodes, node values, and edge values, used to record the total number of times each current capture checkpoint appears within a set of current temporary spatiotemporal spaces and the co-occurrence count of each pair of current capture checkpoints; nodes represent current capture checkpoints; node values ​​represent the total number of times each current capture checkpoint appears within a set of current temporary spatiotemporal spaces; and edge values ​​represent the co-occurrence count of two current capture checkpoints.

[0093] For example, taking the target person's image file as file B, file B exists in a temporary spacetime, which contains 3 checkpoints, namely C B1 C B2 C B3 The temporary spacetime checkpoints are associated using the graph connection method, as shown in the graph below. Figure 3 As shown, this can represent a temporary spacetime checkpoint C. B1 C B2 C B3 The association relationships are such that the association count between checkpoints is 1, and each checkpoint appears 1 time. Further, all the checkpoints in temporary spacetime of file B are connected together in the form of a connection graph, resulting in a new connection graph, such as... Figure 4 As shown, checkpoint C B1 It appeared a total of 7 times, 5 of which were related to checkpoint C. B3 They appeared simultaneously, with 5 of them being related to checkpoint C. B2 At the same time, other checkpoints are similar.

[0094] The association between different capture checkpoints within a set of temporary spatiotemporal spaces can be represented by the association probability corresponding to each capture checkpoint. The association probability corresponding to each capture checkpoint can be determined based on the total number of times each capture checkpoint appears within the set of temporary spatiotemporal spaces and the co-occurrence frequency of each capture checkpoint pair. Here, the association probability corresponding to each capture checkpoint can include the association probability between each capture checkpoint and other capture checkpoints in the set besides itself. The association probability between each capture checkpoint and other capture checkpoints can be the probability of other capture checkpoints appearing given that each capture checkpoint appears within a temporary spatiotemporal space in the target portrait file.

[0095] Optionally, the correlation probability matrix can be used to store the correlation probability between each capture point and other capture points. Here, the horizontal axis of the correlation probability matrix records a set of capture points in the target portrait file, and the vertical axis records the probability that each capture point in the set appears simultaneously with other capture points outside of that set.

[0096] For example, such as Figure 4 The shown bayonet C B3 There are three temporary spacetime checkpoints associated with it, namely C B1 C B2 C B4 Therefore, when a temporary spacetime bottleneck C appears... B3 At that time, checkpoint C appeared. B1 The probability of encountering checkpoint C is 5 / 9. B2 The probability of encountering checkpoint C is 7 / 9. B4The probability is 1 / 9. Calculate in the same way to obtain the correlation probability matrix between temporary spacetime checkpoints. The correlation probability matrix between temporary spacetime checkpoints of file B can be shown in Table 1.

[0097] Table 1

[0098] <![CDATA[C B1 ]]> <![CDATA[C B2 ]]> <![CDATA[C B3 ]]> <![CDATA[C B4 ]]> <![CDATA[C B5 ]]> <![CDATA[C B6 ]]> <![CDATA[C B7 ]]> <![CDATA[C B8 ]]> <![CDATA[C B1 ]]> 1 5 / 7 5 / 7 0 0 0 0 0 <![CDATA[C B2 ]]> 5 / 9 1 7 / 9 4 / 9 4 / 9 2 / 9 0 0 <![CDATA[C B3 ]]> 5 / 9 7 / 9 1 1 / 9 0 0 0 0 <![CDATA[C B4 ]]> 0 4 / 6 1 / 6 1 4 / 6 1 / 6 0 0 <![CDATA[C B5 ]]> 0 4 / 7 0 4 / 7 1 2 / 7 0 0 <![CDATA[C B6 ]]> 0 2 / 4 0 1 / 4 2 / 4 1 0 0 <![CDATA[C B7 ]]> 0 0 0 0 0 0 1 0 <![CDATA[C B8 ]]> 0 0 0 0 0 0 3 / 5 1

[0099] In this embodiment, based on the association probability corresponding to the capture checkpoint in each temporary spacetime of a set of temporary spacetimes, the target temporary spacetime with a missed checkpoint can be determined.

[0100] In this embodiment, the target temporary space with a missing checkpoint is determined by the association probability corresponding to the capture checkpoint under each temporary space in a set of temporary spaces, which can improve the efficiency of determining the target temporary space and the missing checkpoint.

[0101] In an exemplary embodiment, based on the association probability corresponding to the capture checkpoint under each temporary temporal space, a set of target temporary temporal spaces containing missed checkpoints is determined, including:

[0102] S51, Perform the following operations on each temporary spacetime as the current temporary spacetime to obtain the target temporary spacetime:

[0103] Identify a set of current capture checkpoints that appear in the current temporary time and space, and a set of candidate capture checkpoints that do not appear in the current temporary time and space.

[0104] Based on the correlation probability between each current capture checkpoint in a set of current capture checkpoints and each candidate capture checkpoint in a set of candidate capture checkpoints, determine the missed pass probability corresponding to each candidate capture checkpoint.

[0105] If there is a missing checkpoint in a set of candidate capture checkpoints with a missing probability greater than or equal to a preset probability threshold, the current temporary spacetime is determined as the target temporary spacetime.

[0106] When identifying a target temporary spacetime with a missing capture point, the probability of any capture point appearing in each temporary spacetime can be determined based on the correlation probability between each capture point and other capture points, and then combined with the existing capture data in each temporary spacetime to determine the probability.

[0107] When determining the target temporary space containing a missed checkpoint based on the association probability corresponding to the capture checkpoint under each temporary space, each temporary space can be used as the current temporary space to perform the following operations to obtain the target temporary space:

[0108] Identify a set of current capture checkpoints that appear in the current temporary time and space, and a set of candidate capture checkpoints that do not appear in the current temporary time and space. Based on the correlation probability between each current capture checkpoint in the set of current capture checkpoints and each candidate capture checkpoint in the set of candidate capture checkpoints, determine the missed pass probability corresponding to each candidate capture checkpoint.

[0109] Optionally, the correlation probabilities between each current capture checkpoint and each candidate capture checkpoint can be multiplied together to obtain the result. Then, the square root of the result is taken based on the number of candidate capture checkpoints to obtain the missed capture probability corresponding to each candidate checkpoint. Alternatively, the correlation probabilities between each current capture checkpoint and each candidate capture checkpoint can be summed to obtain the sum. Then, the average of the sum is calculated based on the number of candidate capture checkpoints to obtain the missed capture probability corresponding to each candidate capture checkpoint.

[0110] If, among a set of candidate capture points, there exists a capture point with a missed capture probability greater than or equal to a preset probability threshold, the current temporary spacetime can be designated as the target temporary spacetime. Here, the preset probability threshold can be a pre-defined probability threshold.

[0111] For example, the probability of a certain capture point appearing in a temporary space-time of a person's portrait file can be as follows:

[0112]

[0113] in, This indicates that a temporary time-space snapshot occurred at camera C in the portrait file. i The probability, C j This indicates the camera capture point that appeared under this temporary time and space. This indicates that the image file was captured by camera C. j At the same time, the capture checkpoint C appeared. i The probability (which can be determined based on the aforementioned correlation probability matrix between temporary spatiotemporal checkpoints) is calculated, where n represents the number of different capture checkpoints appearing in that temporary spatiotemporal space for the portrait file, and θ is a very small constant (to prevent the probability of occurrence being 0). Based on the correlation probability matrix between temporary spatiotemporal checkpoints for each portrait file, the probability of occurrence of the capture checkpoint corresponding to each temporary spatiotemporal space is calculated. If the probability of occurrence of a capture checkpoint in a certain temporary spatiotemporal space is greater than a threshold, but there is no capture data for that capture checkpoint, it is considered that there may be a missed capture in that temporary spatiotemporal space.

[0114] This embodiment calculates the probability of a missed shot at any capture point under each temporary spatiotemporal space by associating probabilities, thereby determining the missed capture point and the corresponding target temporary spatiotemporal space. This can improve the accuracy of determining the missed capture point, thereby improving the efficiency of processing portrait files.

[0115] In one exemplary embodiment, before recalling the unfiled captured images that match the missed checkpoints in the target temporary spatiotemporal context to the target portrait file, the method further includes:

[0116] S61, filter out the uncollected images from the uncollected image set whose capture time is within the time range corresponding to the target temporary space-time, and obtain a set of candidate capture images;

[0117] S62, if there is a target capture image in a set of candidate capture images with a similarity to the target portrait file greater than or equal to the first similarity threshold, the target capture image is determined as an unfiled capture image that matches the missing checkpoint in the target's temporary spatiotemporal context. The first similarity threshold is negatively correlated with the missing probability corresponding to the missing checkpoint in the target's temporary spatiotemporal context.

[0118] For any unfiled captured images that may correspond to missed checkpoints, in this embodiment, the unfiled captured images can be filtered based on the time range of the target's temporary spatiotemporal space. The unfiled captured images that need to be recalled can be determined by calculating the similarity between the images in the target's portrait file and the filtered unfiled captured images.

[0119] In this embodiment, a set of candidate capture images can be obtained by filtering out the uncollected capture images whose capture time is within the time range corresponding to the target temporary spacetime.

[0120] Optionally, considering that unfiled captured images are those that are discarded when determining target portrait files because their similarity does not meet the first target threshold for file aggregation, the first similarity threshold used to determine whether the similarity between a group of candidate captured images and the target portrait file meets the requirements can be preset to a value smaller than the first target threshold.

[0121] Optionally, the first similarity threshold can be negatively correlated with the probability of a missed checkpoint under the temporary spatiotemporal conditions of the target; that is, the greater the probability of a missed checkpoint, the smaller the first similarity threshold.

[0122] For example, the first similarity threshold could be Where S1 represents the similarity threshold of the archive aggregation (i.e., the aforementioned first target threshold), and α is a constant parameter. For the checkpoint C that may have gaps in this temporary spacetime. i The probability of.

[0123] In this embodiment, if there is a target capture image in a set of candidate capture images whose similarity to the target portrait file is greater than or equal to the first similarity threshold, the target capture image can be identified as an unfiled capture image that matches the missing checkpoint in the target's temporary spatiotemporal context.

[0124] For example, taking unfiled captured images as discarded data, based on the possible missed checkpoints in each temporary time frame of each file, we search for all discarded images from checkpoints that may have missed files within that temporary time frame. We then compare the view similarity of the selected discarded images with the files. If the similarity is greater than... Then a recall of the defective chips will be carried out.

[0125] In this embodiment, by reducing the clustering threshold based on the probability of missing data at the missed data checkpoint, the accuracy of recalling images that have not been captured can be improved.

[0126] In one exemplary embodiment, after recalling the unfiled captured images that match the missing checkpoints in the target temporary spatiotemporal context to the target portrait file, the method further includes:

[0127] S71, Select from a set of portrait files the portrait files containing the corresponding capture time within the time range corresponding to the target temporary time and space, and the corresponding capture checkpoint matching the missing checkpoint under the target temporary time and space, to obtain the candidate portrait files;

[0128] S72, if the similarity between the target image file and the candidate image file is greater than or equal to the second similarity threshold, it is determined that the target image file and the candidate image file are allowed to be merged. The second similarity threshold is negatively correlated with the probability of missing files corresponding to the missing file checkpoint under the target temporary spatiotemporal conditions.

[0129] For the possible portrait files corresponding to the missed checkpoints, in this embodiment, the portrait files can be screened based on the time range of the target temporary space-time and whether there is captured data from the missed checkpoints. By calculating the similarity between the target portrait file and other screened portrait files, it can be determined whether the files need to be merged.

[0130] In this embodiment, a candidate image file can be obtained by filtering out image files from a set of image files that contain image capture data whose corresponding capture time is within the time range corresponding to the target temporary spacetime and whose corresponding capture checkpoint matches the missing checkpoint under the target temporary spacetime.

[0131] Optionally, considering that the candidate image file may be an image file that was separately archived when determining the target image file because the similarity did not meet the second target threshold for file merging, the second similarity threshold used to determine whether the similarity between the candidate image file and the target image file meets the requirements can be preset to a value smaller than the second target threshold.

[0132] Optionally, the second similarity threshold can be negatively correlated with the probability of a missed checkpoint under the temporary spatiotemporal conditions of the target; that is, the greater the probability of a missed checkpoint, the smaller the second similarity threshold.

[0133] Optionally, when determining the second similarity threshold, the probability of the checkpoint with the highest probability of missing a checkpoint under the target temporary spatiotemporal conditions can be selected as the condition for determining the second similarity threshold. Alternatively, multiple probabilities of missing checkpoints can be selected as the condition for determining the second similarity threshold, such as selecting the probabilities of missing checkpoints with the highest, second highest, and third highest probabilities of missing checkpoints.

[0134] For example, the second similarity threshold could be Where S2 represents the similarity threshold for file merging (i.e., the second target threshold), and β1, β2, β3 are constant parameters, where β1 > β2 > β3. This represents the probability of a file being missed at the checkpoint with the highest probability of such a miss within the file. This represents the probability of missing data at the checkpoint with the second highest probability of missing data in the file. This represents the probability of missing data at the checkpoint with the third highest probability of missing data in the file.

[0135] In this embodiment, after recalling the unfiled captured images that match the missing checkpoints in the target temporary space-time to the target portrait file, if the similarity between the target portrait file and the candidate portrait file is greater than or equal to the second similarity threshold, it can be determined that the target portrait file and the candidate portrait file are allowed to be merged.

[0136] Optionally, the similarity between the target portrait file and the candidate portrait file can be determined by calculating the similarity between all captured images in the target portrait file, including the recalled unfiled captured images, and all captured images in the candidate portrait file. The method for determining whether unfiled captured images need to be recalled can be the same as in the aforementioned embodiments, and will not be repeated here.

[0137] For example, based on the potential missing checkpoints in each temporary time frame of each file, find all file data for those checkpoints that might have missing files within that temporary time frame. Compare the view similarity of the selected files with the actual file. If the similarity is greater than... Then the files will be merged.

[0138] This embodiment improves the accuracy of file merging by reducing the merging threshold based on the probability of missing files at the missing file checkpoint.

[0139] The following explanation illustrates the method for processing portrait archives in this application embodiment, using optional examples. In this optional example, the preset area is a designated area, the association relationship is an association probability matrix, and the target capture trajectory is the archive trajectory information.

[0140] In related technologies, due to differences in capture distance, capture angle, and capture equipment, the similarity of images of the same person may not meet the criteria for file aggregation, resulting in multiple files for the same person and making it impossible to completely reconstruct the portrait trajectory. To address this, this optional example provides a method for recalling and merging unusable portrait images based on the temporary spatiotemporal checkpoint relationships within the archives. This method mines the temporary spatiotemporal trajectory of all portrait archives in a specified area based on checkpoint names, checkpoint AOI relationships, and checkpoint distances. Based on the relationships between capture checkpoints within each archive's temporary spatiotemporal space, it identifies checkpoints that may have missed images in each temporary spatiotemporal space. By reasonably lowering the aggregation and merging thresholds, unusable images and archives are recalled and merged, effectively improving the portrait aggregation effect.

[0141] like Figure 5 As shown, the process of processing portrait files in this optional example may include the following steps:

[0142] Step S502: Mine all temporary spacetime traces of all portrait files in the specified area for the past N days (generally N>=7).

[0143] Extract all capture data of facial images in a specified area over the past N days, sort each facial image according to the capture time, and obtain the image trajectory information;

[0144] Extract the Chinese name of each capture checkpoint, segment the Chinese name into words and extract keywords to obtain the updated file trajectory information;

[0145] Based on archival trajectory information, and in accordance with temporary spatiotemporal division rules, we explore the temporary spatiotemporal context of each portrait archive.

[0146] Step S504: Based on the temporary spatiotemporal space within the archives, explore the correlation between temporary spatiotemporal capture checkpoints.

[0147] By exploring the correlation between temporary snapshot checkpoints in the archive, the correlation probability matrix between different snapshot checkpoints in a person's portrait archive is calculated.

[0148] Based on the correlation probability matrix between temporary spatiotemporal capture checkpoints in each portrait file, calculate the probability of occurrence of a capture checkpoint in a temporary spatiotemporal space.

[0149] If the probability of a certain capture point appearing in a certain temporary time and space is greater than a threshold, but there is no corresponding capture data in that temporary time and space, it is considered that there may be a missed capture point in that temporary time and space.

[0150] Step S506: Recall discarded films and merge archives based on temporary spatiotemporal gaps.

[0151] For each portrait file, under each temporary time and space, there may be missed capture points, and all unusable image data of the capture points that may have missed captures within that temporary time and space range are searched.

[0152] The filtered rejected images are compared with the portrait file in terms of visual similarity. If the similarity is greater than 100%, the image is considered to be similar to the portrait file. Then, a recall of the defective chips will be conducted;

[0153] For each portrait file, under each temporary time and space, there may be missed capture points, and search for all file data of capture points that may have missed captures within that temporary time and space.

[0154] The similarity of other selected profiles to the current profile is compared. If the similarity is greater than 1, the profile is considered similar to the other profiles. Then the files will be merged.

[0155] This optional example helps to identify the relationships between capture points in different temporary time and space within a portrait archive by determining the correlation between capture points and capture points that may have been missed in each temporary time and space. This assists in the recall of discarded images and the merging of archives, and can effectively improve the recall rate of portrait archives while ensuring accuracy.

[0156] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0157] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM (Read-Only Memory) / RAM (Random Access Memory), magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of this application.

[0158] According to another aspect of the embodiments of this application, an apparatus for implementing the above-described method for processing portrait archives is also provided. Figure 6 This is a structural block diagram of an optional facial image processing apparatus according to an embodiment of this application, such as... Figure 6 As shown, the device may include:

[0159] Extraction unit 602 is used to acquire a first captured image of a target scene captured by a target shooting device, and to perform occlusion processing on a first image region in the first captured image to obtain a first occluded image, wherein the first image region corresponds to a first scene region in the target scene;

[0160] The first determining unit 604, connected to the extraction unit 602, is used to determine the number of target-type objects currently residing in the first image area based on the first captured image or the first occlusion image, and to obtain the first target number.

[0161] The recall unit 606 is connected to the first determination unit 604 and is used to determine the number of target type objects currently in the first image area based on the first captured image or the first occlusion image, so as to obtain the first target number.

[0162] The closing unit 608, connected to the recall unit 606, is used to occlude the second image region in the second captured image to obtain a second occluded image. The second captured image is a captured image of the target scene by the target capturing device after the first captured image is captured.

[0163] It should be noted that the extraction unit 602 in this embodiment can be used to execute the above step S202, the first determination unit 604 in this embodiment can be used to execute the above step S204, the recall unit 606 in this embodiment can be used to execute the above step S206, and the file merging unit 608 in this embodiment can be used to execute the above step S208.

[0164] Through the above modules, a set of capture data from a target portrait file within a preset time period is extracted from a set of portrait files corresponding to a preset area, resulting in a set of temporary spatiotemporal data within the target portrait file. Each temporary spatiotemporal data segment contains a continuous sequence of capture data associated with corresponding spatiotemporal information. The spatiotemporal information of each capture data segment records the capture checkpoint and capture time corresponding to that segment. Based on the correlation between different capture checkpoints within a set of temporary spatiotemporal data, a target temporary spatiotemporal data segment with missing capture checkpoints is identified. The missing capture checkpoints within the target temporary spatiotemporal data segment are predicted checkpoints that lack corresponding capture data within the target temporary spatiotemporal data segment. The target temporary spatiotemporal data segment is then linked to the target temporary spatiotemporal data segment. Incomplete snapshots captured under spatiotemporal conditions are recalled to the target portrait archive to update it. These incomplete snapshots do not belong to any portrait archive within a set of portrait archives. If a candidate portrait archive exists within a set of portrait archives that temporarily matches the target, and the similarity between the target and candidate portrait archives determines that merging the target and candidate portrait archives is permissible, then the target and candidate portrait archives are merged. This addresses the problem of low recall rates in portrait archive processing methods due to low accuracy in recognizing images of the same object, thus improving the recall rate of portrait archive merging.

[0165] In one exemplary embodiment, the extraction unit includes:

[0166] The sorting module is used to sort a set of captured data according to the corresponding capture time to obtain the target capture trajectory corresponding to the target portrait file;

[0167] The first extraction module is used to perform temporary spatiotemporal extraction on the target capture trajectory to obtain a set of temporary spatiotemporal data.

[0168] In one exemplary embodiment, the extraction unit includes:

[0169] The second extraction module is used to extract each segment of continuous capture data that meets the preset extraction conditions from a set of capture data to obtain a set of temporary spatiotemporal data.

[0170] The preset extraction conditions include: the distance between any two adjacent capture data points corresponding to capture checkpoints is less than or equal to a first distance; or there is an intersection between the preset information surfaces of any two adjacent capture data points corresponding to capture checkpoints; or there is an intersection between the keywords contained in the checkpoint identifiers of any two adjacent capture data points corresponding to capture checkpoints. The preset information surface is the information surface corresponding to the second distance; the time difference between any two adjacent capture data points corresponding to capture times is less than or equal to a first time threshold, and the total capture duration is greater than or equal to a second time threshold; and it does not belong to any temporary spacetime within the target portrait file.

[0171] In one exemplary embodiment, the second extraction module includes:

[0172] The sorting submodule is used to sort a set of captured data according to the corresponding capture time to obtain the target capture trajectory corresponding to the target portrait file;

[0173] The first extraction submodule is used to extract keywords from the checkpoint identifier of each capture data in a set of capture data, and obtain the keywords contained in the checkpoint identifier of each capture data.

[0174] The update submodule is used to update each capture data in the target capture trajectory with the keywords contained in the checkpoint identifier of the capture checkpoint corresponding to each capture data, so as to obtain the updated capture trajectory. Each capture data in the updated capture trajectory is used to record the capture checkpoint corresponding to each capture data, the keywords contained in the checkpoint identifier of the corresponding capture checkpoint, and the capture time.

[0175] The second extraction submodule is used to extract temporary spatiotemporal data from the updated capture trajectory to obtain a set of temporary spatiotemporal data.

[0176] In one exemplary embodiment, the first determining unit includes:

[0177] The first determining module is used to determine the total number of times each capture checkpoint in a set of capture checkpoints appears in a set of temporary time and space and the co-occurrence number of each capture checkpoint pair in a set of capture checkpoints. Here, a set of capture checkpoints refers to all capture checkpoints in a set of temporary time and space, and the co-occurrence number of each capture checkpoint pair is the total number of times the two capture checkpoints in each capture checkpoint pair appear in the same temporary time and space.

[0178] The second determining module is used to determine the association probability corresponding to each capture checkpoint based on the total number of times each capture checkpoint appears in a set of temporary spatiotemporal spaces and the co-occurrence number of each capture checkpoint pair. The association probability corresponding to each capture checkpoint includes the association probability between each capture checkpoint and other capture checkpoints in the set of capture checkpoints besides each capture checkpoint. The association probability between each capture checkpoint and other capture checkpoints is the probability of other capture checkpoints appearing when each capture checkpoint appears in a temporary spatiotemporal space within the target portrait file.

[0179] The third determining module is used to determine the target temporary space with a missed checkpoint in a set of temporary spaces based on the association probability corresponding to the capture checkpoint in each temporary space in a set of temporary spaces.

[0180] In one exemplary embodiment, the third determining module includes:

[0181] The execution submodule is used to perform the following operations on each temporary spacetime as the current temporary spacetime to obtain the target temporary spacetime:

[0182] Identify a set of current capture checkpoints that appear in the current temporary time and space, and a set of candidate capture checkpoints that do not appear in the current temporary time and space.

[0183] Based on the correlation probability between each current capture checkpoint in a set of current capture checkpoints and each candidate capture checkpoint in a set of candidate capture checkpoints, determine the missed pass probability corresponding to each candidate capture checkpoint.

[0184] If there is a missing checkpoint in a set of candidate capture checkpoints with a missing probability greater than or equal to a preset probability threshold, the current temporary spacetime is determined as the target temporary spacetime.

[0185] In one exemplary embodiment, the above-described apparatus further includes:

[0186] The first filtering unit is used to filter out unfiled snapshots whose capture time is within the time range corresponding to the target temporary space-time before recalling the unfiled snapshots that match the missing checkpoints in the target temporary space-time to the target portrait file, and obtain a set of candidate snapshots.

[0187] The second determining unit is used to determine the target capture image as an unfiled capture image that matches the missing checkpoint in the target temporary spatiotemporal context when there is a target capture image in a set of candidate capture images with a similarity greater than or equal to the target portrait file. The first similarity threshold is negatively correlated with the missing probability corresponding to the missing checkpoint in the target temporary spatiotemporal context.

[0188] In one exemplary embodiment, the above-described apparatus further includes:

[0189] The second filtering unit is used to, after recalling the unfiled captured images that match the missing checkpoints under the target temporary time and space to the target portrait archive, filter out the portrait archives containing the corresponding capture time within the time range corresponding to the target temporary time and space, and the corresponding capture checkpoints that match the missing checkpoints under the target temporary time and space, to obtain the candidate portrait archives.

[0190] The third determining unit is used to determine whether the target image file and the candidate image file are allowed to be merged when the similarity between the target image file and the candidate image file is greater than or equal to the second similarity threshold. The second similarity threshold is negatively correlated with the probability of missing files corresponding to the missing file checkpoint under the target temporary spatiotemporal conditions.

[0191] It should be noted that the examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments. It should also be noted that the above modules, as part of a device, can operate in environments such as... Figure 1 The hardware environment shown can be implemented through software or hardware, and the hardware environment includes the network environment.

[0192] According to another aspect of the embodiments of this application, a storage medium is also provided. Optionally, in this embodiment, the storage medium can be used to execute program code for any of the above-described methods for processing portrait files in the embodiments of this application.

[0193] Optionally, in this embodiment, the storage medium may be located on at least one of the network devices in the network shown in the above embodiment.

[0194] Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps:

[0195] S1, Temporary spatiotemporal extraction is performed on a set of capture data of a target portrait file in a set of portrait files corresponding to a preset area within a preset time period to obtain a set of temporary spatiotemporal data within the target portrait file. Among them, a temporary spatiotemporal data contains a segment of continuous capture data associated with corresponding spatiotemporal information. The spatiotemporal information of a capture data is used to record the capture checkpoint and capture time corresponding to a capture data.

[0196] S2, based on the correlation between different capture checkpoints under a set of temporary time and space, determine a target temporary time and space with missing checkpoints in a set of temporary time and space, wherein the missing checkpoints under the target temporary time and space are the predicted checkpoints that are missing corresponding capture data under the target temporary time and space;

[0197] S3, recall the unfiled captured images that match the missing checkpoints in the target temporary space-time to the target portrait file to update the target portrait file. Among them, the unfiled captured images do not belong to any portrait file in a set of portrait files.

[0198] S4. If a candidate image file that temporarily matches the target image file exists in a set of image files, and the target image file and the candidate image file are allowed to be merged based on the similarity between the target image file and the candidate image file, then merge the target image file and the candidate image file.

[0199] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments, and will not be repeated in this embodiment.

[0200] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing program code, such as USB flash drives, ROMs, RAMs, portable hard drives, magnetic disks, or optical disks.

[0201] According to another aspect of the embodiments of this application, an electronic device for implementing the above-described method for processing portrait archives is also provided. The electronic device may be a server, a terminal, or a combination thereof.

[0202] Figure 7 This is a structural block diagram of an optional electronic device according to an embodiment of this application, such as... Figure 7 As shown, it includes a processor 702, a communication interface 704, a memory 706, and a communication bus 708. The processor 702, communication interface 704, and memory 706 communicate with each other via the communication bus 708.

[0203] Memory 706 is used to store computer programs;

[0204] When processor 706 executes a computer program stored in memory 706, it performs the following steps:

[0205] S1, Temporary spatiotemporal extraction is performed on a set of capture data of a target portrait file in a set of portrait files corresponding to a preset area within a preset time period to obtain a set of temporary spatiotemporal data within the target portrait file. Among them, a temporary spatiotemporal data contains a segment of continuous capture data associated with corresponding spatiotemporal information. The spatiotemporal information of a capture data is used to record the capture checkpoint and capture time corresponding to a capture data.

[0206] S2, based on the correlation between different capture checkpoints under a set of temporary time and space, determine a target temporary time and space with missing checkpoints in a set of temporary time and space, wherein the missing checkpoints under the target temporary time and space are the predicted checkpoints that are missing corresponding capture data under the target temporary time and space;

[0207] S3, recall the unfiled captured images that match the missing checkpoints in the target temporary space-time to the target portrait file to update the target portrait file. Among them, the unfiled captured images do not belong to any portrait file in a set of portrait files.

[0208] S4. If a candidate image file that temporarily matches the target image file exists in a set of image files, and the target image file and the candidate image file are allowed to be merged based on the similarity between the target image file and the candidate image file, then merge the target image file and the candidate image file.

[0209] Optionally, in this embodiment, the communication bus can be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. This communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, Figure 7 The symbol is represented by a single thick line, but this does not indicate that there is only one bus or one type of bus. The communication interface is used for communication between the aforementioned electronic device and other devices.

[0210] The memory may include RAM, or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0211] As an example, the memory 706 described above may include, but is not limited to, the extraction unit 602, the first determination unit 604, the recall unit 606, and the file merging unit 608 from the aforementioned image file processing device. Furthermore, it may include, but is not limited to, other module units from the aforementioned image file processing device, which will not be elaborated upon in this example.

[0212] The processors mentioned above can be general-purpose processors, including but not limited to: CPU (Central Processing Unit), NP (Network Processor), etc.; they can also be DSP (Digital Signal Processor), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0213] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments, and will not be repeated here.

[0214] Those skilled in the art will understand that Figure 7 The structure shown is for illustrative purposes only. The device that implements the above-described method for processing facial images can be a terminal device, such as a smartphone (e.g., an Android phone, an iOS phone), a tablet computer, a PDA, a mobile internet device (MID), a tablet computer, or a PAD. Figure 7 This does not limit the structure of the aforementioned electronic device. For example, the electronic device may also include components that are more... Figure 7 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 7 The different configurations shown.

[0215] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, ROM, RAM, disk or optical disk, etc.

[0216] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0217] If the integrated units in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in the aforementioned 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 one or more computer devices (which may be personal computers, servers, or network devices, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0218] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0219] In the several embodiments provided in this application, it should be understood that the disclosed client can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection between units or modules, and may be electrical or other forms.

[0220] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the solution provided in this embodiment, depending on actual needs.

[0221] 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.

[0222] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method of processing a portrait archive, characterized by, include: Temporary spatiotemporal extraction is performed on a set of capture data of a target portrait file in a set of portrait files corresponding to a preset area within a preset time period to obtain a set of temporary spatiotemporal data within the target portrait file. Among them, a temporary spatiotemporal data contains a segment of continuous capture data associated with corresponding spatiotemporal information. The spatiotemporal information of a capture data is used to record the capture checkpoint and capture time corresponding to the capture data. Based on the correlation between different capture checkpoints in the set of temporary spatiotemporal spaces, a target temporary spatiotemporal space with missing checkpoints is determined. The missing checkpoints in the target temporary spatiotemporal space are predicted checkpoints that lack corresponding capture data in the target temporary spatiotemporal space. The correlation between different capture checkpoints in the set of temporary spatiotemporal spaces includes the correlation probability corresponding to each capture checkpoint. The correlation probability corresponding to each capture checkpoint is determined based on the total number of times each capture checkpoint appears in the set of temporary spatiotemporal spaces and the co-occurrence count of each capture checkpoint pair. The target temporary spatiotemporal space with missing checkpoints in the set of temporary spatiotemporal spaces is determined based on the correlation probability corresponding to the capture checkpoints in each temporary spatiotemporal space within the set of temporary spatiotemporal spaces. Unfiled captured images that match the missing checkpoints in the target temporary space-time are recalled to the target portrait file to update the target portrait file. The unfiled captured images do not belong to any portrait file in the set of portrait files. If a candidate image file that temporarily matches the target image file exists in the set of image files, and the target image file and the candidate image file are allowed to be merged based on the similarity between the target image file and the candidate image file, then the target image file and the candidate image file are merged.

2. The method of claim 1, wherein, The step of temporarily extracting a set of captured data from a set of target portrait files within a preset time period from a set of portrait files corresponding to a preset area, to obtain a set of temporary spatiotemporal data within the target portrait file, includes: The set of captured data is sorted according to the corresponding capture time to obtain the target capture trajectory corresponding to the target portrait file; Temporary spatiotemporal extraction is performed on the target capture trajectory to obtain the set of temporary spatiotemporal data.

3. The method of claim 1, wherein, The step of temporarily extracting a set of captured data from a set of target portrait files within a preset time period from a set of portrait files corresponding to a preset area, to obtain a set of temporary spatiotemporal data within the target portrait file, includes: Extract each segment of continuous capture data that meets the preset extraction conditions from the set of capture data to obtain the set of temporary spatiotemporal data; The preset extraction conditions include: the distance between any two adjacent capture data points corresponding to capture checkpoints is less than or equal to a first distance; or there is an intersection between the preset information surfaces of any two adjacent capture data points corresponding to capture checkpoints; or there is an intersection between the keywords contained in the checkpoint identifiers of any two adjacent capture data points corresponding to capture checkpoints. The preset information surface is the information surface corresponding to a second distance; the time difference between any two adjacent capture data points corresponding to capture times is less than or equal to a first time threshold, and the total capture duration is greater than or equal to a second time threshold; and it does not belong to any temporary spacetime within the target portrait file.

4. The method of claim 3, wherein, The step of extracting each segment of continuous capture data that meets preset extraction conditions from the set of capture data to obtain the set of temporary spatiotemporal data includes: The set of captured data is sorted according to the corresponding capture time to obtain the target capture trajectory corresponding to the target portrait file; Keyword extraction is performed on the checkpoint identifier corresponding to each capture data in the set of capture data to obtain the keywords contained in the checkpoint identifier corresponding to each capture data. The keywords contained in the checkpoint identifier of the checkpoint corresponding to each capture data are used to update each capture data in the target capture trajectory to obtain an updated capture trajectory. The capture data in the updated capture trajectory is used to record the capture checkpoint corresponding to each capture data, the keywords contained in the checkpoint identifier of the corresponding capture checkpoint, and the capture time. Temporary spatiotemporal extraction is performed on the updated capture trajectory to obtain the set of temporary spatiotemporal data.

5. The method according to any one of claims 1 to 4, characterized in that, The step of determining the target temporary temporal space with a missed capture point in the set of temporary temporal spaces based on the correlation between different capture points under the set of temporary temporal spaces includes: Determine the total number of times each capture checkpoint in a set of capture checkpoints appears in the set of temporary spatiotemporal spaces and the co-occurrence count of each capture checkpoint pair in the set of capture checkpoints, wherein the set of capture checkpoints refers to all capture checkpoints in the set of temporary spatiotemporal spaces, and the co-occurrence count of each capture checkpoint pair is the total number of times the two capture checkpoints in each capture checkpoint pair appear in the same temporary spatiotemporal space; Based on the total number of times each capture point appears in the set of temporary spatiotemporal spaces and the co-occurrence number of each capture point pair, the association probability corresponding to each capture point is determined. The association probability corresponding to each capture point includes the association probability between each capture point and other capture points in the set of capture points besides each capture point. The association probability between each capture point and the other capture points is the probability of the other capture points appearing when each capture point appears in a temporary spatiotemporal space within the target portrait file. Based on the association probability corresponding to the capture checkpoint in each temporary spacetime of the set of temporary spacetimes, the target temporary spacetime with a missed checkpoint is determined.

6. The method according to claim 5, characterized in that, The step of determining the target temporary temporal space containing a missed checkpoint in the group of temporary temporal spaces based on the association probability corresponding to the capture checkpoint under each temporary temporal space includes: Perform the following operations on each of the aforementioned temporary spacetimes as the current temporary spacetime to obtain the target temporary spacetime: Determine a set of current capture checkpoints that appear in the current temporary time and space, and a set of candidate capture checkpoints that do not appear in the current temporary time and space; Based on the correlation probability between each current capture checkpoint in the set of current capture checkpoints and each candidate capture checkpoint in the set of candidate capture checkpoints, determine the missed pass probability corresponding to each candidate capture checkpoint; If there is a missed capture point in the set of candidate capture points with a missed capture probability greater than or equal to a preset probability threshold, the current temporary spacetime is determined as the target temporary spacetime.

7. The method according to claim 5, characterized in that, Before recalling the unfiled captured images that match the missing checkpoints in the target's temporary spatiotemporal context to the target's portrait file, the method further includes: A set of candidate images is obtained by filtering out unfiled images from the set of unfiled images whose capture time is within the time range corresponding to the temporary spatiotemporal space of the target. If, in the set of candidate captured images, there is a target captured image with a similarity to the target portrait file greater than or equal to a first similarity threshold, the target captured image is determined as an unfiled captured image that matches the missing checkpoint in the target's temporary spatiotemporal context. The first similarity threshold is negatively correlated with the missing probability corresponding to the missing checkpoint in the target's temporary spatiotemporal context.

8. The method according to claim 5, characterized in that, After recalling the unfiled captured images that match the missing checkpoints in the target's temporary spatiotemporal context to the target's portrait file, the method further includes: The candidate image file is obtained by selecting the image file containing the corresponding capture time within the time range corresponding to the target temporary space-time and the corresponding capture checkpoint matching the missing checkpoint under the target temporary space-time. If the similarity between the target image file and the candidate image file is greater than or equal to a second similarity threshold, it is determined that the target image file and the candidate image file are allowed to be merged, wherein the second similarity threshold is negatively correlated with the probability of missing files corresponding to the missing file checkpoint under the target temporary spatiotemporal conditions.

9. A device for processing portrait archives, characterized in that, include: The extraction unit is used to temporarily extract a set of capture data of a target portrait file in a set of portrait files corresponding to a preset area within a preset time period, to obtain a set of temporary spatiotemporal data within the target portrait file. Among them, a temporary spatiotemporal data contains a segment of continuous capture data associated with corresponding spatiotemporal information. The spatiotemporal information of a capture data is used to record the capture checkpoint and capture time corresponding to the capture data. The first determining unit is used to determine, based on the correlation between different capture checkpoints in the set of temporary temporal spaces, a target temporary temporal space containing a missing checkpoint, wherein the missing checkpoint in the target temporary temporal space is a predicted checkpoint that lacks corresponding capture data in the target temporary temporal space; the correlation between different capture checkpoints in the set of temporary temporal spaces includes the correlation probability corresponding to each capture checkpoint, and the correlation probability corresponding to each capture checkpoint is determined based on the total number of times each capture checkpoint appears in the set of temporary temporal spaces and the co-occurrence number of each capture checkpoint pair; the target temporary temporal space containing a missing checkpoint is determined based on the correlation probability corresponding to the capture checkpoint in each temporary temporal space in the set of temporary temporal spaces. The recall unit is used to recall the unfiled captured images that match the missing checkpoints in the target temporary time and space to the target portrait file, so as to update the target portrait file. The unfiled captured images do not belong to any of the captured images in the set of portrait files. The file merging unit is used to merge the target image file and the candidate image file when there is a candidate image file that temporarily matches the target image file in the set of image files, and the target image file and the candidate image file are allowed to be merged based on the similarity between the target image file and the candidate image file.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein the program, when executed, performs the method of any one of claims 1 to 8.

11. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to execute the method of any one of claims 1 to 8 through the computer program.

Images