Image file gathering method and device and electronic equipment

Abstract

The invention provides an image file gathering method and device and electronic equipment, and relates to the technical field of image processing, and the method comprises the steps: obtaining a human face file identifier corresponding to a human body file in a human body file list; obtaining a target face file associated with each face file identifier in a face file list; the human body archive list and the human face archive list are obtained by gathering images including human bodies and human faces and collected at snapshot points in a monitoring system within a preset time period; and on the basis of the number of the target face archives, maintaining the human body archives to human body archive lists with different credibility degrees. According to the method and the device, when the files of the non-face snapshot images are gathered, the files can be gathered based on the human body file lists with different credibility degrees, so that the accuracy of the file gathering of the non-face snapshot images is improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to an image aggregation method, device and electronic equipment. Background Art

[0002] Portrait archiving is one of the core businesses of the current security industry. Its purpose is to extract and compare the features of portrait snapshots captured by video surveillance equipment, and archive the portrait snapshots belonging to the same person together to form a person track. Currently, the technology of using faces for portrait archiving is very mature, but in some cases, the face cannot be well captured due to back, side or lighting problems. At this time, human body recognition technology is needed to complete the person track.

[0003] In the related technology, facial files are first clustered based on facial features to obtain facial files, and then the body features corresponding to the faces are clustered to obtain body files. When there are new non-face captured images, the features of the non-face captured images are matched with the body features in all body files. If the match is successful, the non-face captured images are archived into the corresponding body files.

[0004] However, the above-mentioned related technologies are affected by characteristics such as clothing and body shape, so the confidence of human body features is lower than that of facial features. The same human body features may correspond to different faces, thereby reducing the accuracy of clustering non-face captured images based on all human body files. Summary of the invention

[0005] In view of the problems existing in the prior art, the embodiments of the present invention provide an image archiving method, device and electronic device.

[0006] The present invention provides an image aggregation method, comprising:

[0007] Get the face file identifier corresponding to the human file in the human file list;

[0008] Obtaining the target face files associated with each face file identifier in the face file list; the body file list and the face file list are both obtained based on the image files including the body and the face collected at each capture point in the monitoring system within a preset time period;

[0009] Based on the number of the target face profiles, each of the human body profiles is maintained in a list of human body profiles with different credibility levels.

[0010] According to an image archiving method provided by the present invention, the method of maintaining each of the human body archives into a list of human body archives with different credibility levels based on the number of the target face archives includes:

[0011] When the number of target face files is equal to one, the identification of the target body file corresponding to the target face file, the identification of the target face file, the body center of mass feature included in the target body file, and the corresponding relationship of the first effective range are added to the trusted body file list; the first effective range is used to characterize the identification of all capture points in the monitoring system.

[0012] According to an image archiving method provided by the present invention, the method of maintaining each of the human body archives into a list of human body archives with different credibility levels based on the number of the target face archives includes:

[0013] When the number of the target facial files is greater than one and it is determined that there are capture points in the same area between the capture points corresponding to the target facial files, the identifiers of the target body files corresponding to all the target facial files, the identifiers of the target facial files, the body centroid features included in the target body files, and the corresponding relationship of the second effective range are added to the untrusted body file list; the second effective range is used to characterize the identifiers of the capture points corresponding to the target body files.

[0014] According to an image file gathering method provided by the present invention, the step of determining that there are capture points in the same area between the capture points corresponding to each target face file includes:

[0015] For each two target face files, based on the directed connected graphs corresponding to the two target face files, determine whether there are the same capture points between the capture points corresponding to one target face file and the capture points corresponding to the other target face file in the two target face files; the directed connected graph is used to represent the personnel trajectory corresponding to the facial centroid features included in the target face files, and each node in the directed connected graph is used to represent the capture point of the facial image corresponding to the facial centroid features included in the target face files;

[0016] In the case where the same capture points exist, it is determined that there are capture points in the same area between the capture points corresponding to the two target face files.

[0017] According to an image aggregation method provided by the present invention, the method further includes:

[0018] In the case where there are no identical capture points, based on the directed connected graphs corresponding to the two target face files, respectively, the shortest path between the capture point corresponding to one target face file and the capture point corresponding to the other target face file in the two target face files is determined; the edge in the directed connected graph is used to represent the difference between the capture time of the facial image corresponding to the facial centroid feature at the first capture point and the capture time of the second capture point, and the node corresponding to the first capture point and the node corresponding to the second capture point are connected to the edge;

[0019] When the length of the shortest path is less than a preset length, it is determined that there are capture points in the same area between the capture points corresponding to the two target face files.

[0020] According to an image archiving method provided by the present invention, the method of maintaining each of the human body archives into a list of human body archives with different credibility levels based on the number of the target face archives includes:

[0021] When the number of the target facial files is greater than one and there are no capture points in the same area between the capture points corresponding to the target facial files, for each target facial file, the identification of the target body file corresponding to the target facial file, the identification of the target facial file, the body center of mass features included in the target body file, and the corresponding relationship of the third effective range are added to the trusted body file list; the third effective range is used to characterize the identification of the capture point corresponding to the target body file.

[0022] According to an image aggregation method provided by the present invention, the method further includes:

[0023] Acquire a human target image without a face collected within the preset time period;

[0024] When it is determined that there is an identifier representing a capture point corresponding to the human target image without a face in the second effective range, matching the human centroid features corresponding to the capture points of the human target image without a face in the untrusted human file list with the human features of the human target image without a face;

[0025] In the case of successful matching, the human target image without a face is discarded.

[0026] According to an image aggregation method provided by the present invention, the method further includes:

[0027] In the event of a match failure, the centroid features of each body in the trusted body file list are traversed, and the currently traversed body centroid feature is matched with the body feature of the faceless body target image; in the event of a match success, the image capture identifier of the faceless body target image is stored in the body file corresponding to the successfully matched body centroid feature.

[0028] The present invention also provides an image aggregation device, comprising:

[0029] A first acquisition unit is used to acquire a face file identifier corresponding to a human body file in a human body file list;

[0030] The second acquisition unit is used to obtain the target face files associated with each face file identifier in the face file list; the human body file list and the face file list are both obtained based on the image files including human bodies and faces collected at each capture point in the monitoring system within a preset time period;

[0031] The file aggregation unit is used to maintain each of the human body files into a list of human body files with different credibility levels based on the number of the target face files.

[0032] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the image archiving method described above is implemented.

[0033] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described image archiving methods.

[0034] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the image archiving method described above is implemented.

[0035] The image archiving method, device and electronic device provided by the present invention obtain the face file identifier corresponding to the human body file in the human body file list, obtain the target face file associated with each face file identifier in the face file list, and maintain each human body file in a human body file list with different credibility levels based on the number of target face files. In this way, when the captured images without faces are aggregated, the images can be aggregated based on the human body file lists with different credibility levels, thereby improving the accuracy of the aggregated images without faces. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0037] Figure 1 This is one of the flowcharts of the image aggregation method provided by an embodiment of the present invention;

[0038] Figure 2 This is a second flow chart of the image aggregation method provided by an embodiment of the present invention;

[0039] Figure 3 This is a third flow chart of the image aggregation method provided by an embodiment of the present invention;

[0040] Figure 4 This is a fourth flow chart of the image aggregation method provided by an embodiment of the present invention;

[0041] Figure 5 This is a fifth flow chart of the image aggregation method provided by an embodiment of the present invention;

[0042] Figure 6 is a structural schematic diagram of an image aggregation device provided by an embodiment of the present invention;

[0043] Figure 7 It is a schematic diagram of the physical structure of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] Combine the following Figure 1-Figure 5 The image archiving method of the present invention is described. The execution subject of the method can be an electronic device such as a computer, a server, a cloud server or a server cluster, or an image archiving device set in the electronic device, and the image archiving device can be implemented by software, hardware or a combination of the two. The method can be applied in a monitoring scenario with multiple monitoring devices, for example, multiple monitoring devices are used to monitor pedestrians on urban roads, and each monitoring device corresponds to a capture point.

[0046] Figure 1is one of the flow charts of the image aggregation method provided by an embodiment of the present invention, such as Figure 1 As shown, the image aggregation method includes the following steps:

[0047] Step 101, obtaining the facial profile identifiers corresponding to the human profiles with the same human centroid features in the human profile list.

[0048] Among them, the human body file list includes at least one human body file, and each human body file includes the correspondence between the human body file identifier, at least one face file identifier, a human body centroid feature and at least one image capture identifier, wherein the human body centroid feature is the feature that is closest to the average distance of the human body features of multiple human body capture images, and the human body centroid feature is used as the representative feature of the human body features of multiple human body capture images; there can be multiple image capture identifiers corresponding to the human body file identifier, and the image capture identifier can be understood as the number of the image captured by the monitoring equipment, etc. For example, if 10,000 images are captured, the 10,000 images are numbered, and the numbers can be 1, 2, 3...10000.

[0049] For example, due to factors such as the capture angle, light, and the fact that the human features corresponding to different faces may be the same, for example, the human features of person 1 include height of 170 cm, thin, and wearing a black suit, and the human features of person 2 include height of 170 cm, thin, and wearing a black suit. As a result, there may be a situation in the human file list where one human file is associated with multiple facial file identifiers. In this case, the facial file identifier corresponding to the human file obtained in the human file list may be one or more.

[0050] Step 102, obtaining the target face files associated with each face file identifier in the face file list; the body file list and the face file list are both obtained based on the image aggregation including the body and the face collected at each capture point in the monitoring system within a preset time period.

[0051] Among them, the preset time period can be set based on needs. For example, the preset time period can be greater than or equal to one week. The personnel trajectory determined based on images including human bodies and faces collected over a period of more than one week can cover most of a person’s life scenes and is more representative.

[0052] In a preset time period, each capture point of the monitoring system can collect multiple images including human bodies and human faces. The human face part in the image can be understood as a human face image that can accurately identify a person, that is, a frontal face image, a side face image, or an image containing the facial features of a person. Based on the facial features in the image, multiple images are feature-gathered to obtain a list of face files. The face file list includes at least one face file. Each face file includes a face file identifier, a face centroid feature, and a corresponding relationship between at least one image capture identifier. Among them, the face centroid feature is the feature with the closest average distance to the face features of multiple face capture images. The face centroid feature is used as a representative feature of the face features of multiple human body capture images. Then, based on the human body features in the multiple images, the files are gathered to obtain a list of human body files. The human body file list includes at least one human body file. Each human body file includes a human body file identifier, at least one face file identifier, a corresponding relationship between the human body centroid feature, and at least one image capture identifier. Since the image includes both facial features and human body features, the face and the human body are associated. The human body here is the human body associated with the face.

[0053] Since different people may correspond to the same human body profile, based on the association between face and body, all human body profiles in the human body profile list can be associated with at least one corresponding facial profile identifier, and the target facial profile corresponding to each facial profile identifier is obtained in the facial profile list. There are one or more target facial profiles, and these target facial profiles can be considered as facial profiles directly associated with the human body profile.

[0054] Step 103: Based on the number of the target face profiles, maintain the human body profiles into a list of human body profiles with different credibility levels.

[0055] The human profile lists of different credibility levels may include a credible human profile list and an uncredible human profile list.

[0056] For example, if there is only one target face file obtained, it means that the human body file and the target face file correspond one-to-one, which means that the human body centroid feature included in the human body file is prominent enough, and the human body file to which the human body centroid feature belongs can be maintained in the trusted human body file list; if the number of target face files obtained is greater than one, it is necessary to determine whether to maintain the human body file in the trusted human body file list or in the untrusted human body file list based on whether there are snapshot points in the same area between the snapshot points corresponding to the target face files.

[0057] The image clustering method provided by the present invention obtains the face file identifier corresponding to the human body file in the human body file list, obtains the target face file associated with each face file identifier in the face file list, and maintains each human body file in a human body file list with different credibility levels based on the number of target face files. In this way, when clustering the captured images without faces, the clustering can be performed based on the human body file lists with different credibility levels, thereby improving the accuracy of clustering the captured images without faces.

[0058] In one embodiment, Figure 2 FIG. 2 is a flow chart of the image aggregation method provided by an embodiment of the present invention. Figure 2 As shown, the above step 103 maintains each of the human body profiles into a list of human body profiles with different credibility levels based on the number of the target face profiles, which can be specifically achieved by the following steps:

[0059] Step 1031, when the number of the target face files is equal to one, the identification of the target body file corresponding to the target face file, the identification of the target face file, the body center of mass feature included in the target body file, and the corresponding relationship of the first effective range are added to the trusted body file list; the first effective range is used to characterize the identification of all capture points in the monitoring system.

[0060] For example, when the number of target face files is equal to one, it means that the human body file corresponds to the target face file one by one, and the human body centroid feature included in the human body file is prominent enough, that is, the human body feature of the person corresponding to the human body centroid feature is prominent enough, and the entire monitoring system has not found that the human body features of other persons are the same as the human body features of the person within the preset time period, and it has uniqueness within the preset time period. At this time, the human body file to which the human body centroid feature belongs can be determined as a valid file, that is, the relevant information of the human body file is stored in the trusted human body file list, and it is considered that in the entire monitoring system, there is a globally trusted association between the human body file and the target face file; specifically, the identifier of the target human body file corresponding to the target face file, the identifier of the target face file, the human body centroid feature and the corresponding relationship of the first effective range are added to the trusted human body file list; wherein, the first effective range is used to characterize the identifier of all capture points in the entire monitoring system, and it is considered that all capture points in the entire monitoring system are globally valid, that is, all capture points in the entire monitoring system are valid and trusted capture points.

[0061] In this embodiment, when the number of target face files is equal to one, the identifier of the target body file corresponding to the target face file, the identifier of the target face file, the correspondence between the body centroid feature and the effective range are added to the trusted body file list, that is, the relevant information of the body file with sufficiently prominent features is added to the trusted body file list, so as to facilitate the subsequent clustering of non-face captured images based on the body file with sufficiently prominent features.

[0062] In one embodiment, Figure 3 FIG. 3 is a flow chart of the image aggregation method provided by an embodiment of the present invention. Figure 3 As shown, the above step 103 maintains each of the human body profiles into a list of human body profiles with different credibility levels based on the number of the target face profiles, which can be specifically achieved by the following steps:

[0063] Step 1032: When the number of the target facial files is greater than one and it is determined that there are capture points in the same area between the capture points corresponding to the target facial files, the identifiers of the target body files corresponding to all the target facial files, the identifiers of each of the target facial files, the body center of mass features included in the target body files, and the corresponding relationship of the second effective range are added to the untrusted body file list; the second effective range is used to characterize the identifiers of the capture points corresponding to each of the target body files.

[0064] For example, when the number of target face files is greater than one, it means that one human body file corresponds to multiple target face files, which means that different persons have the same human body file. The entire monitoring system finds that the human body features of other persons are the same as the human body features of the person within a preset time period. At this time, it is necessary to determine whether there are capture points in the same area between the capture points corresponding to the target face files, that is, to determine whether there are the same capture points or adjacent capture points between the capture points corresponding to the target face files. When it is determined that there are the same capture points or adjacent capture points between the capture points corresponding to the target face files, it is said that It is clear that different people with the same body features have tracks in the same area, that is, there are different people with very similar body shapes and clothing in the same area, so it is impossible to uniquely identify a person based on the body file. The confidence of the body file is low and cannot be used as a reference for judging the identity of the person. At this time, the relevant information of the body file is stored in the untrusted body file list; specifically, the identification of the same target body file corresponding to all target face files, the identification of each target face file, the body center of mass features included in the target body file, and the corresponding relationship of the second effective range are added to the untrusted body file list. The second effective range is used to represent the identification of the capture point corresponding to each target body file. If the capture point represented by the second effective range collects a body image that matches the body center of mass features included in the target body file, then the confidence of this body image is very low and cannot be used as a basis for confirming the identity of the person.

[0065] In this embodiment, when the number of target facial files is greater than one and it is determined that there are snapshot points in the same area between the snapshot points corresponding to the target facial files, the identifiers of the same target human body file corresponding to all target facial files, the identifiers of each target facial file, the human body centroid features and the corresponding relationship of the second effective range are added to the untrusted human body file list, that is, the relevant information of the human body file that cannot be uniquely identified as a person is added to the untrusted human body file list, so that when it is determined that the human body capture image without a face matches the human body centroid features included in the human body file in the untrusted human body file list, the human body capture image without a face is discarded and not used as a basis for person identity confirmation, thereby improving the accuracy of identity confirmation of the human body capture image without a face.

[0066] In one embodiment, the determination in the above step 1032 that there are capture points in the same area between the capture points corresponding to the target face files can be implemented in the following way:

[0067] For every two target face files, based on the directed connected graphs corresponding to the two target face files, determine whether there are the same capture points between the capture points corresponding to one target face file and the capture points corresponding to the other target face file in the two target face files; when there are the same capture points, determine whether there are capture points in the same area between the capture points corresponding to the two target face files.

[0068] Among them, the directed connected graph is used to represent the personnel trajectory corresponding to the facial centroid features included in the target facial file, and each node in the directed connected graph is used to represent the capture point of the facial image corresponding to the facial centroid features included in the target facial file.

[0069] For example, face aggregation can be used to generate the trajectory of each person. According to the spatiotemporal connectivity of the personnel trajectory, a directed connected graph of the capture points corresponding to each person's trajectory is constructed. Each node in the directed connected graph is used to represent a capture point, and the edge between two nodes is used to represent the shortest passing time of a person from a corresponding capture point to another capture point. The method of constructing the directed connected graph of the capture points corresponding to each personnel trajectory is the same, so that a directed connected graph of the entire monitoring system can be constructed.

[0070] Among them, the calculation method of the shortest passing time is: traverse the capture information corresponding to each image capture identifier in all image capture identifiers included in the face file, each capture information includes a capture point and a corresponding capture time, calculate the time difference between each capture time corresponding to a capture point and each capture time corresponding to another capture point, and determine the minimum time difference as the shortest passing time from the capture point to another capture point. In this way, the capture points can be sorted based on the time sequence, and the identifiers of each capture point are used as the attributes of the nodes of the directed connected graph, and the shortest passing time is used as the attribute of the edge connecting the two nodes. The specific directed connected graph example is as follows:

[0071] Face F1: A--10min→B--5min→C---20minvD

[0072] Face F2: B--1min→A--5minvD---10min→E

[0073] Face F3: B--3min→E--8min→C---15min→A

[0074] Face FN:………………

[0075] Taking the directed connectivity graph of face F1 as an example, the shortest passing time for the person corresponding to face F1 to run from capture point A to capture point B is 10 minutes, the shortest passing time from capture point B to capture point C is 5 minutes, and the shortest passing time from capture point C to capture point D is 20 minutes. The meanings of the directed connectivity graphs of other human faces are the same as those of the directed connectivity graph of face F1, and the present invention will not be repeated here.

[0076] For example, since each face file can correspond to a directed connected graph, when the number of target face files is greater than one, for each two target face files, based on the directed connected graphs corresponding to the two target face files, it can be determined whether there are the same capture points between the capture points corresponding to one target face file and the capture points corresponding to the other target face file in the two target face files. In the case of the same capture points, it means that the trajectories of the persons corresponding to the two target face files have overlapping parts; at this time, it is determined that there are capture points in the same area between the capture points corresponding to each target face file. For example, the faces corresponding to the two target face files are face F1 and face F2, respectively. It can be seen from the directed connected graph of face F1 and the directed connected graph of face F2 that the persons corresponding to face F1 and the persons corresponding to face F2 both have trajectories at capture point A and capture point B, so it is determined that there are capture points in the same area between the capture points corresponding to the target face file of face F1 and the capture points corresponding to the target face file of face F2.

[0077] In this embodiment, based on the directed connected graph corresponding to each target facial file, it can be determined whether there are the same capture points between the capture points corresponding to each target facial file, so as to judge whether there are overlapping trajectories of the persons corresponding to the target facial files. The determination method is fast and accurate.

[0078] In one embodiment, the determination in step 1032 that there are capture points in the same area between the capture points corresponding to the target face files may be implemented in the following manner:

[0079] In the case that there are no identical capture points, based on the directed connected graphs corresponding to the two target face files, the shortest path between the capture points corresponding to one target face file and the capture points corresponding to the other target face file in the two target face files is determined; in the case that the length of the shortest path is less than a preset length, it is determined that there are capture points in the same area between the capture points corresponding to the two target face files.

[0080] The edge in the directed connected graph is used to represent the difference between the capture time of the face image corresponding to the face centroid feature at the first capture point and the capture time of the second capture point, and the node corresponding to the first capture point and the node corresponding to the second capture point are connected to the edge. The difference can be understood as the above-mentioned shortest passing time.

[0081] For example, when there are no identical snapshot points, it means that the trajectories of the persons corresponding to the two target face files have no overlapping parts. At this time, it is necessary to determine whether the trajectories of the persons corresponding to the two target face files are adjacent. The specific method for determining whether they are adjacent is: based on the directed connected graphs corresponding to the two target face files, a shortest path search algorithm is used to determine the shortest path between the snapshot point corresponding to one target face file and the snapshot point corresponding to the other target face file in the two target face files, and the length of the shortest path is compared with a preset length. When it is determined that the length of the shortest path is less than the preset length, it means that the trajectories of the persons corresponding to the two target face files are adjacent. At this time, it is determined that there are snapshot points in the same area between the snapshot points corresponding to the two target face files; when it is determined that the length of the shortest path is greater than or equal to the preset length, it is determined that the trajectories of the persons corresponding to the two target face files are not adjacent, that is, it is determined that there are no snapshot points in the same area between the snapshot points corresponding to the two target face files. Among them, the preset length can be determined based on the total number of capture points in the monitoring system and the density of the capture points. The smaller the preset length, the higher the accuracy of human body gathering. For example, the preset length can be 10 minutes.

[0082] In this embodiment, when there are no identical capture points, it is further determined whether the persons corresponding to the two target face files have adjacent trajectories based on the directed connected graphs corresponding to the two target face files, and the determination method is fast and accurate.

[0083] In one embodiment, Figure 4 FIG. 4 is a flow chart of the image aggregation method provided by an embodiment of the present invention. Figure 4 As shown, the above step 103 maintains each of the human body profiles into a list of human body profiles with different credibility levels based on the number of the target face profiles, which can be specifically achieved by the following steps:

[0084] Step 1033: When the number of the target facial files is greater than one and there are no capture points in the same area between the capture points corresponding to the target facial files, for each target facial file, the identifier of the target body file corresponding to the target facial file, the identifier of the target facial file, the body center of mass feature included in the target body file, and the corresponding relationship of the third effective range are added to the trusted body file list; the third effective range is used to characterize the identifier of the capture point corresponding to the target body file.

[0085] For example, when the number of target face files is greater than one, and there are no capture points in the same area between the capture points corresponding to the target face files, it means that although different persons have the same body file, the entire monitoring system has not found other persons with the same body features as the person in the same area during the preset time period. That is to say, within the preset time period, different persons with the same body file have their own trajectories in different areas, that is, there are no different persons with very high similarity in body shape and clothing in the same area, so a person can be determined based on the body file, and the body file can be used as a reference for judging the identity of the person. At this time, the relevant information of the body file is stored in the trusted body file list; specifically, since the same body file is associated with multiple target face files, the target body file associated with the multiple target face files is copied into the same number of target body files as the target face files. For example, if there are 3 target face files associated with the target body file, the target body file is copied into 3 target body files. At this time, it is considered that each copied target body file has a local trusted association with the target face file associated with it. It should be noted here that if the capture points corresponding to the copied target body file are the same as the capture points corresponding to the associated target face file, it means that the capture points corresponding to each copied target body file are different.

[0086] For each target human body profile, the identification of the target human body profile, the identification of the target facial profile, the human body centroid features included in the target human body profile, and the corresponding relationship of the third effective range are added to the trusted human body profile list. Each copied target human body profile corresponds to a third effective range. The third effective range is used to characterize the identification of the snapshot point corresponding to the target human body profile. If the snapshot point represented by the third effective range captures a human body image that matches the human body centroid features included in the target human body profile, the confidence of this human body image is relatively high and can be used as a basis for confirming the identity of the person.

[0087] In this embodiment, when the number of target facial files is greater than one and there are no capture points in the same area between the capture points corresponding to the target facial files, the target body file is copied based on the number of target facial files, and for each target body file, the correspondence between the target body file identifier, the target facial file identifier, the body centroid feature and the third effective range is added to the trusted body file list, that is, the relevant information that can determine the body file of a person is added to the trusted body file list, so that when it is determined that the non-face capture image matches the body centroid feature included in the body file in the trusted body file list, the non-face capture image can be used as the basis for confirming the identity of the person.

[0088] After the face and body association and archiving process is completed, a list of trusted body files and a list of untrusted body files are generated, that is, the process of maintaining all body files within the preset time period into a list of body files with different levels of credibility is completed. Next, the human body captured images without face association collected within the preset time period can be processed based on the trusted body file list and the untrusted body file list.

[0089] In one embodiment, Figure 5 FIG. 5 is a flowchart of the image aggregation method provided by an embodiment of the present invention. Figure 5 As shown, after step 103, the image aggregation method further includes the following steps:

[0090] Step 104: Acquire a human target image without a face collected within the preset time period.

[0091] For example, the images captured by each capture point in the monitoring system within a preset time period may include both faces and bodies, or may only include bodies without faces, so it is possible to obtain human target images without faces captured within the preset time period; human target images without faces can be understood as images that cannot be used to identify specific people based on face recognition but contain human bodies; for example, human target images without faces are images with no faces at all and only human bodies, or images that include human bodies and partial faces but cannot be used to identify specific people based on face recognition, or images that include the entire back of a person, etc.

[0092] Step 105: When it is determined that there is an identification of the capture point corresponding to the human target image without a face in the second effective range, the human center of mass features corresponding to the capture point of the human target image without a face in the untrusted human file list are matched with the human features of the human target image without a face.

[0093] Step 106: If the matching is successful, discard the human target image without a face.

[0094] For example, for a human target image without a face, first obtain the capture point of the human target image without a face, and then search in the untrusted human archive list to see whether there is a capture point of the human target image without a face. When the capture point of the human target image without a face is found in the untrusted human archive list, match the human features of the human target image without a face with the human centroid features of the image collected at the capture point of the human target image without a face. When the match is successful, it means that the human target image without a face is also untrustworthy and cannot be used as a basis for determining the identity of a person. In this case, the human target image without a face is discarded.

[0095] In this embodiment, when it is determined that there is an identification of the capture point corresponding to the human target image without a face in the second effective range, the human center of mass features corresponding to the capture points of the human target image without a face in the untrusted human file list are matched with the human features of the human target image without a face. If the match is successful, it means that the human target image without a face is also untrustworthy, and the human target image without a face is discarded, thereby avoiding storing the image capture identification corresponding to the untrustworthy human target image without a face in the corresponding human file. In other words, it is avoided that people with similar body shape and clothing are grouped into the same human file, thereby improving the accuracy of grouping the human target image without a face.

[0096] In one embodiment, if Figure 5 As shown, the image aggregation method also includes the following steps:

[0097] Step 107, in the event of a matching failure, traverse the centroid features of each body in the trusted body file list, match the currently traversed body centroid feature with the body feature of the faceless body target image, and in the event of a matching success, store the image capture identifier of the faceless body target image in the body file corresponding to the successfully matched body centroid feature.

[0098] For example, when the matching fails, it means that the human target image without a face is credible, that is, the human features of the human target image without a face are prominent enough, or, there are no other human features similar to the human features of the human target image without a face in the same area. At this time, the human centroid features in the credible human file list are traversed, and the currently traversed human centroid features are matched with the human features of the human target image without a face. When the matching is successful, it means that the human features of the human target image without a face are also prominent enough, or, there are no other human features similar to the human features of the human target image without a face in the same area, which means that the human target image without a face is also credible, and the image capture identifier of the human target image without a face is stored in the human file corresponding to the human centroid feature of the successful match.

[0099] In this embodiment, in the event of a match failure in the untrusted human body file list, the centroid features of each human body in the trusted human body file list continue to be traversed, and the centroid features of the currently traversed human body are matched with the human body features of the target human body image without a face. In the event of a successful match, the image capture identifier of the target human body image without a face is stored in the human body file corresponding to the centroid feature of the successfully matched human body, thereby achieving accurate archiving of credible target human body images without a face.

[0100] The image gathering device provided by the present invention is described below. The image gathering device described below and the image gathering method described above can be referred to each other.

[0101] Figure 6 is a schematic diagram of the structure of an image aggregation device provided by an embodiment of the present invention. Figure 6 As shown, the image gathering device 600 includes a first acquisition unit 601, a second acquisition unit 602 and a gathering unit 603; wherein:

[0102] The first acquisition unit 601 is used to acquire the face file identifier corresponding to the human body file in the human body file list;

[0103] The second acquisition unit 602 is used to obtain the target face files associated with each face file identifier in the face file list; the human body file list and the face file list are both obtained based on the image files including human bodies and faces collected at each capture point in the monitoring system within a preset time period;

[0104] The file aggregation unit 603 is used to maintain each of the human body files into a list of human body files with different credibility levels based on the number of the target face files.

[0105] The image archiving device provided by the present invention obtains the face file identifier corresponding to the human body file in the human body file list, obtains the target face file associated with each face file identifier in the face file list, and maintains each human body file in a human body file list with different credibility levels based on the number of target face files. In this way, when the captured images without faces are aggregated, the images can be aggregated based on the human body file lists with different credibility levels, thereby improving the accuracy of the aggregated images without faces.

[0106] Based on any of the above embodiments, the aggregate unit 603 is specifically used for:

[0107] When the number of target face files is equal to one, the identification of the target body file corresponding to the target face file, the identification of the target face file, the body center of mass feature included in the target body file, and the corresponding relationship of the first effective range are added to the trusted body file list; the first effective range is used to characterize the identification of all capture points in the monitoring system.

[0108] Based on any of the above embodiments, the aggregate unit 603 is specifically used for:

[0109] When the number of the target facial files is greater than one and it is determined that there are capture points in the same area between the capture points corresponding to the target facial files, the identifiers of the target body files corresponding to all the target facial files, the identifiers of the target facial files, the body centroid features included in the target body files, and the corresponding relationship of the second effective range are added to the untrusted body file list; the second effective range is used to characterize the identifiers of the capture points corresponding to the target body files.

[0110] Based on any of the above embodiments, the aggregate unit 603 is further specifically configured to:

[0111] For each two target face files, based on the directed connected graphs corresponding to the two target face files, determine whether there are the same capture points between the capture points corresponding to one target face file and the capture points corresponding to the other target face file in the two target face files; the directed connected graph is used to represent the personnel trajectory corresponding to the facial centroid features included in the target face files, and each node in the directed connected graph is used to represent the capture point of the facial image corresponding to the facial centroid features included in the target face files;

[0112] In the case where the same capture points exist, it is determined that there are capture points in the same area between the capture points corresponding to the two target face files.

[0113] Based on any of the above embodiments, the aggregate unit 603 is further specifically configured to:

[0114] In the case where there are no identical capture points, based on the directed connected graphs corresponding to the two target face files, respectively, the shortest path between the capture point corresponding to one target face file and the capture point corresponding to the other target face file in the two target face files is determined; the edge in the directed connected graph is used to represent the difference between the capture time of the facial image corresponding to the facial centroid feature at the first capture point and the capture time of the second capture point, and the node corresponding to the first capture point and the node corresponding to the second capture point are connected to the edge;

[0115] When the length of the shortest path is less than a preset length, it is determined that there are capture points in the same area between the capture points corresponding to the two target face files.

[0116] Based on any of the above embodiments, the aggregate unit 603 is specifically used for:

[0117] When the number of the target facial files is greater than one and there are no capture points in the same area between the capture points corresponding to the target facial files, for each target facial file, the identification of the target body file corresponding to the target facial file, the identification of the target facial file, the body center of mass features included in the target body file, and the corresponding relationship of the third effective range are added to the trusted body file list; the third effective range is used to characterize the identification of the capture point corresponding to the target body file.

[0118] Based on any of the above embodiments, the image aggregation device 600 further includes:

[0119] A third acquisition unit is used to acquire a human target image without a face collected within the preset time period;

[0120] a first matching unit, configured to match the human centroid features corresponding to the snapshot points of the human target image without a face in the untrusted human file list with the human features of the human target image without a face, when it is determined that there is an identifier representing the snapshot point corresponding to the human target image without a face in the second effective range;

[0121] The first matching unit is further configured to discard the human target image without a face if the matching is successful.

[0122] Based on any of the above embodiments, the image aggregation device 600 further includes:

[0123] The second matching unit is used to traverse the centroid features of each body in the trusted body file list when the matching fails, match the centroid feature of the body currently traversed with the body feature of the target body image without a face, and when the matching is successful, store the image capture identifier of the target body image without a face in the body file corresponding to the centroid feature of the body with successful matching.

[0124] Figure 7 is a schematic diagram of the physical structure of an electronic device provided by an embodiment of the present invention, such as Figure 7As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730 and a communication bus 740, wherein the processor 710, the communication interface 720 and the memory 730 communicate with each other via the communication bus 740. The processor 710 may call the logic instructions in the memory 730 to execute the image archiving method, which includes: obtaining a face file identifier corresponding to a human file in a human file list;

[0125] Obtaining the target face files associated with each face file identifier in the face file list; the body file list and the face file list are both obtained based on the image files including the body and the face collected at each capture point in the monitoring system within a preset time period;

[0126] Based on the number of the target face profiles, each of the human body profiles is maintained in a list of human body profiles with different credibility levels.

[0127] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0128] On the other hand, the present invention further provides a computer program product, the computer program product includes a computer program, the computer program can be stored in a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the image archiving method provided by the above methods, the method includes: obtaining a face file identifier corresponding to a human file in a human file list;

[0129] Obtaining the target face files associated with each face file identifier in the face file list; the body file list and the face file list are both obtained based on the image files including the body and the face collected at each capture point in the monitoring system within a preset time period;

[0130] Based on the number of the target face profiles, each of the human body profiles is maintained in a list of human body profiles with different credibility levels.

[0131] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to execute the image archiving method provided by the above methods, the method comprising: obtaining a face profile identifier corresponding to a human profile in a human profile list;

[0132] Obtaining the target face files associated with each face file identifier in the face file list; the body file list and the face file list are both obtained based on the image files including the body and the face collected at each capture point in the monitoring system within a preset time period;

[0133] Based on the number of the target face profiles, each of the human body profiles is maintained in a list of human body profiles with different credibility levels.

[0134] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0135] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0136] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An image aggregation method, characterized in that: include: Get the face file identifier corresponding to the human file in the human file list; Obtaining the target face files associated with each face file identifier in the face file list; the body file list and the face file list are both obtained based on the image files including the body and the face collected at each capture point in the monitoring system within a preset time period; Based on the number of the target face profiles, each of the human body profiles is maintained in a list of human body profiles with different credibility levels.

2. The image aggregation method according to claim 1, characterized in that: The maintaining of each of the human body profiles into a list of human body profiles with different credibility levels based on the number of the target face profiles includes: When the number of target face files is equal to one, the identification of the target body file corresponding to the target face file, the identification of the target face file, the body center of mass feature included in the target body file, and the corresponding relationship of the first effective range are added to the trusted body file list; the first effective range is used to characterize the identification of all capture points in the monitoring system.

3. The image aggregation method according to claim 2, characterized in that: The maintaining of each of the human body profiles into a list of human body profiles with different credibility levels based on the number of the target face profiles includes: When the number of the target facial files is greater than one and it is determined that there are capture points in the same area between the capture points corresponding to the target facial files, the identifiers of the target body files corresponding to all the target facial files, the identifiers of the target facial files, the body centroid features included in the target body files, and the corresponding relationship of the second effective range are added to the untrusted body file list; the second effective range is used to characterize the identifiers of the capture points corresponding to the target body files.

4. The image aggregation method according to claim 3, characterized in that: The step of determining that there are capture points in the same area between the capture points corresponding to the target face files includes: For each two target face files, based on the directed connected graphs corresponding to the two target face files, determine whether there are the same capture points between the capture points corresponding to one target face file and the capture points corresponding to the other target face file in the two target face files; the directed connected graph is used to represent the personnel trajectory corresponding to the facial centroid features included in the target face files, and each node in the directed connected graph is used to represent the capture point of the facial image corresponding to the facial centroid features included in the target face files; In the case where the same capture points exist, it is determined that there are capture points in the same area between the capture points corresponding to the two target face files.

5. The image aggregation method according to claim 4, characterized in that: The method further comprises: In the case where there are no identical capture points, based on the directed connected graphs corresponding to the two target face files, respectively, the shortest path between the capture point corresponding to one target face file and the capture point corresponding to the other target face file in the two target face files is determined; the edge in the directed connected graph is used to represent the difference between the capture time of the facial image corresponding to the facial centroid feature at the first capture point and the capture time of the second capture point, and the node corresponding to the first capture point and the node corresponding to the second capture point are connected to the edge; When the length of the shortest path is less than a preset length, it is determined that there are capture points in the same area between the capture points corresponding to the two target face files.

6. The image aggregation method according to claim 3, characterized in that: The maintaining of each of the human body profiles into a list of human body profiles with different credibility levels based on the number of the target face profiles includes: When the number of the target facial files is greater than one and there are no capture points in the same area between the capture points corresponding to the target facial files, for each target facial file, the identification of the target body file corresponding to the target facial file, the identification of the target facial file, the body center of mass features included in the target body file, and the corresponding relationship of the third effective range are added to the trusted body file list; the third effective range is used to characterize the identification of the capture point corresponding to the target body file.

7. The image aggregation method according to claim 6, characterized in that: The method further comprises: Acquire a human target image without a face collected within the preset time period; When it is determined that there is an identifier representing a capture point corresponding to the human target image without a face in the second effective range, matching the human centroid features corresponding to the capture points of the human target image without a face in the untrusted human file list with the human features of the human target image without a face; In the case of successful matching, the human target image without a face is discarded.

8. The image aggregation method according to claim 7, characterized in that: The method further comprises: In the event of a match failure, the centroid features of each body in the trusted body file list are traversed, and the currently traversed body centroid feature is matched with the body feature of the faceless body target image; in the event of a match success, the image capture identifier of the faceless body target image is stored in the body file corresponding to the successfully matched body centroid feature.

9. An image aggregation device, characterized in that: include: A first acquisition unit is used to acquire a face file identifier corresponding to a human body file in a human body file list; The second acquisition unit is used to obtain the target face files associated with each face file identifier in the face file list; the human body file list and the face file list are both obtained based on the image files including human bodies and faces collected at each capture point in the monitoring system within a preset time period; The file aggregation unit is used to maintain each of the human body files into a list of human body files with different credibility levels based on the number of the target face files.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the image archiving method as described in any one of claims 1 to 8 is implemented.

Images