Abnormal intrusion monitoring method and system for smart communities

Abstract

This application relates to the field of smart community monitoring and discloses a method and system for monitoring abnormal intrusions in smart communities. The method includes: determining the key frames of multiple surveillance videos corresponding to a target person, the time sequence of the key frames of the multiple surveillance videos, and the time span between the key frames of the multiple surveillance videos; using an event recognition model, determining the event types and event monitoring indicators corresponding to the key frames of the multiple surveillance videos; constructing a monitoring event graph corresponding to the target person based on the event types and event monitoring indicators corresponding to the key frames of the multiple surveillance videos; determining the abnormal intrusion probability value of the target person based on the monitoring event graph, and when the abnormal intrusion probability value of the target person is greater than or equal to a preset abnormal intrusion probability value, issuing an abnormal intrusion prompt message corresponding to the target person. This application can improve the accuracy of abnormal intrusion monitoring in smart communities and reduce the verification burden on security personnel.

Description

Technical Field

[0001] The present application relates to the field of smart community monitoring technology, and more specifically, to a method and system for monitoring abnormal intrusions in smart communities. Background Art

[0002] Current smart community security systems primarily rely on a combination of traditional video surveillance and manual inspections. In this model, fixed cameras typically provide real-time monitoring of various areas within the community. Furthermore, the response process to intrusion incidents relies on manual review. Once the monitoring system detects an anomaly, security personnel immediately conduct a review and take appropriate action.

[0003] However, this community security system still faces significant technical bottlenecks in detecting unusual intrusions. First, traditional fixed cameras are limited by their physical installation locations, easily creating blind spots in complex areas such as community corners and green belts, preventing full coverage. Second, during the intrusion response process, the time lag between anomaly triggering and alarm confirmation often exceeds 30 seconds, making it difficult to meet the real-time handling requirements of emergencies. Furthermore, community monitoring of unusual human activity primarily relies on intelligent security systems based on facial recognition, but these systems suffer from the following core flaws: a limited target recognition range, making it incapable of identifying unregistered suspicious individuals; insufficient adaptability to dynamic scenarios, and sensitivity to lighting conditions, obstructions, and changes in posture, resulting in a false alarm rate exceeding 40% at night or in complex environments. This not only increases the verification burden on security personnel but can also lead to safety hazards due to missed alerts. Summary of the Invention

[0004] The purpose of this application is to provide a method and system for monitoring abnormal intrusions in smart communities, which solves the technical problem of poor abnormal intrusion monitoring effects in existing smart communities and achieves the technical effect of improving the accuracy of abnormal intrusion monitoring in smart communities.

[0005] An embodiment of the present application provides a method for monitoring abnormal intrusions in a smart community, the method comprising: obtaining multiple surveillance videos corresponding to a target person captured by multiple cameras, and determining the key frames of the multiple surveillance videos corresponding to the target person, the time sequence of the key frames of the multiple surveillance videos, and the time span between the key frames of the multiple surveillance videos; determining, through an event recognition model, event types and event monitoring indicators corresponding to the key frames of the multiple surveillance videos; constructing a monitoring event graph corresponding to the target person based on the event types and event monitoring indicators corresponding to the key frames of the multiple surveillance videos; wherein the monitoring event graph comprises event nodes and edges connecting the event nodes, the attributes of the event nodes include event types and event monitoring indicators, the edges connecting the event nodes connect the event nodes according to the time sequence of the key frames of the multiple surveillance videos, and the attributes of the edges connecting the event nodes include the time span between the event nodes corresponding to the key frames of the surveillance videos; determining, through an abnormal intrusion detection model, the abnormal intrusion probability value of the target person according to the monitoring event graph, and when the abnormal intrusion probability value of the target person is greater than or equal to a preset abnormal intrusion probability value, issuing abnormal intrusion prompt information corresponding to the target person; wherein the abnormal intrusion detection model is a graph neural network model.

[0006] In one possible implementation, the method includes: obtaining regional scenes corresponding to key frames of multiple surveillance videos, and obtaining the maximum movement time, minimum movement time, and normal movement time between regional scenes corresponding to key frames of multiple surveillance videos that are adjacent in time sequence, and adding the maximum movement time, minimum movement time, and normal movement time between regional scenes corresponding to key frames of multiple surveillance videos in the attributes of the edges connecting event nodes; and determining the abnormal intrusion probability value of the target person according to the monitoring event graph through an abnormal intrusion detection model.

[0007] In another possible implementation, the method includes: determining monitoring event graphs at moments corresponding to key frames of multiple monitoring videos, determining the abnormal intrusion probability value of the target person corresponding to each monitoring event graph through an abnormal intrusion detection model, and determining the cumulative value of the abnormal intrusion probability value; when the cumulative value of the abnormal intrusion probability value of the target person is greater than or equal to the cumulative value of the preset abnormal intrusion probability value, issuing abnormal intrusion prompt information corresponding to the target person.

[0008] In another possible implementation, the method includes: obtaining the regional scenes to which the key frames of multiple surveillance videos respectively belong, and obtaining the intrusion probability weight value corresponding to each regional scene, determining the sum of the products of the abnormal intrusion probability values ​​and the intrusion probability weight values ​​corresponding to the regional scenes to which the key frames of the multiple surveillance videos respectively belong, as the cumulative value of the abnormal intrusion probability values; when the cumulative value of the abnormal intrusion probability value of the target person is greater than or equal to the cumulative value of the preset abnormal intrusion probability value, issuing abnormal intrusion prompt information corresponding to the target person.

[0009] In another possible implementation, the method includes: obtaining the regional scenes to which the key frames of multiple surveillance videos respectively belong, and obtaining the key frame number threshold and time span threshold of the surveillance video corresponding to each regional scene; determining the number of key frames of the surveillance video in each regional scene; when the number of key frames of the surveillance video corresponding to the regional scene is less than the key frame number threshold of the surveillance video, or when the time span of the key frames of the surveillance video corresponding to the regional scene is less than the preset time span threshold, deleting the event node corresponding to the key frame of the surveillance video corresponding to the regional scene in the monitoring event graph.

[0010] In another possible implementation, the method includes: obtaining the time periods to which the key frames of multiple surveillance videos belong, and obtaining the time period adjustment factors corresponding to the time periods to which the key frames of multiple surveillance videos belong; determining the product of the key frame number threshold and the time period adjustment factor of the surveillance video corresponding to each regional scene to adjust the key frame number threshold; determining the time span threshold and the time period adjustment factor of the surveillance video corresponding to each regional scene to adjust the time period adjustment factor.

[0011] In another possible implementation, the method includes: determining multiple associated persons of a target person and constructing monitoring event graphs corresponding to each of the multiple associated persons; and determining, using an abnormal intrusion detection model, a group abnormal intrusion probability value corresponding to the target person and the multiple associated persons based on the monitoring event graphs corresponding to the target person and the multiple associated persons. When the group abnormal intrusion probability value corresponding to the target person and the multiple associated persons is greater than or equal to a preset group abnormal intrusion probability value, issuing a group abnormal intrusion prompt message corresponding to the target person and the multiple associated persons.

[0012] In another possible implementation, multiple associated persons of a target person are determined, including: constructing monitoring event graphs corresponding to the multiple persons respectively, and determining the event nodes corresponding to the same area scenes in the monitoring event graphs of each person and the target person respectively; determining the similarity of the event nodes corresponding to the same area scenes of each person and the target person respectively, and determining the sum of the similarities of all event nodes of each person and the target person as the person association degree; when the person association degree of a first person is greater than or equal to a preset person association degree, the first person is regarded as an associated person of the target person.

[0013] In another possible implementation, the method includes: obtaining a similarity adjustment factor corresponding to each regional scene, and determining the product of the similarity of the event nodes corresponding to each person and the target person in the same regional scene and the similarity adjustment factor as the person association.

[0014] An embodiment of the present application also provides an abnormal intrusion monitoring system for a smart community, comprising a unit for executing any of the methods described above.

[0015] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0016] An embodiment of the present application provides a method for monitoring abnormal intrusions in a smart community, the method comprising: obtaining multiple surveillance videos corresponding to a target person captured by multiple cameras, and determining the key frames of the multiple surveillance videos corresponding to the target person, the time sequence of the key frames of the multiple surveillance videos, and the time span between the key frames of the multiple surveillance videos; determining, through an event recognition model, the event types and event monitoring indicators corresponding to the key frames of the multiple surveillance videos; constructing a monitoring event graph corresponding to the target person based on the event types and event monitoring indicators corresponding to the key frames of the multiple surveillance videos; wherein the monitoring event graph comprises event nodes and edges connecting the event nodes, the attributes of the event nodes include event types and event monitoring indicators, the edges connecting the event nodes connect the event nodes according to the time sequence of the key frames of the multiple surveillance videos, and the attributes of the edges connecting the event nodes include the time span between the event nodes corresponding to the key frames of the surveillance videos; determining, through an abnormal intrusion detection model, the abnormal intrusion probability value of the target person according to the monitoring event graph, and when the abnormal intrusion probability value of the target person is greater than or equal to a preset abnormal intrusion probability value, issuing abnormal intrusion prompt information corresponding to the target person; wherein the abnormal intrusion detection model is a graph neural network model. The embodiment of the present application can construct a monitoring event graph of the target person, and combine it with a graph neural network to identify the monitoring event graph, thereby achieving accurate identification of the target person's activity pattern in the community, improving the accuracy of identifying abnormal intrusions of the target person in the community, and improving the abnormal intrusion identification effect of the smart community. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0018] Figure 1A flowchart of a first abnormal intrusion monitoring method for a smart community provided in an embodiment of the present application;

[0019] Figure 2 A flowchart of a second abnormal intrusion monitoring method for a smart community provided in an embodiment of the present application;

[0020] Figure 3 A flowchart of a third abnormal intrusion monitoring method for a smart community provided in an embodiment of the present application;

[0021] Figure 4 A flowchart of a fourth abnormal intrusion monitoring method for a smart community provided in an embodiment of the present application;

[0022] Figure 5 A flowchart of a fifth abnormal intrusion monitoring method for a smart community provided in an embodiment of the present application;

[0023] Figure 6 A flowchart of a sixth abnormal intrusion monitoring method for a smart community provided in an embodiment of the present application;

[0024] Figure 7 A schematic diagram of the logical structure of an abnormal intrusion monitoring system for a smart community provided in an embodiment of the present application. DETAILED DESCRIPTION

[0025] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0026] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0027] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0028] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0029] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0030] The current community security system has the following core flaws: the target recognition range is limited and it cannot identify unregistered suspicious persons; it lacks adaptability to dynamic scenes and is sensitive to lighting conditions, obstructions and posture changes, resulting in a false alarm rate of over 40% at night or in complex environments. This not only increases the verification burden on security personnel, but may also cause safety hazards due to missed reports.

[0031] Based on the above reasons, an embodiment of the present application provides a method for abnormal intrusion monitoring in a smart community, the method comprising: obtaining multiple surveillance videos corresponding to a target person shot by multiple cameras respectively, and determining the key frames of the multiple surveillance videos corresponding to the target person, the time sequence of the key frames of the multiple surveillance videos, and the time span between the key frames of the multiple surveillance videos; through an event recognition model, determining the event types and event monitoring indicators corresponding to the key frames of the multiple surveillance videos respectively according to the key frames of the multiple surveillance videos; constructing a monitoring event graph corresponding to the target person according to the event types and event monitoring indicators corresponding to the key frames of the multiple surveillance videos; wherein the monitoring event graph includes event nodes and edges connecting the event nodes, the attributes of the event nodes include event types and event monitoring indicators, the edges connecting the event nodes connect the event nodes according to the time sequence of the key frames of the multiple surveillance videos, and the attributes of the edges connecting the event nodes include the time span between the event nodes corresponding to the key frames of the surveillance videos; through an abnormal intrusion detection model, determining the abnormal intrusion probability value of the target person according to the monitoring event graph, and when the abnormal intrusion probability value of the target person is greater than or equal to the preset abnormal intrusion probability value, issuing abnormal intrusion prompt information corresponding to the target person; wherein the abnormal intrusion detection model is a graph neural network model. The embodiment of the present application can construct a monitoring event graph of the target person, and combine it with a graph neural network to identify the monitoring event graph, thereby achieving accurate identification of the target person's activity pattern in the community, improving the accuracy of identifying abnormal intrusions of the target person in the community, and improving the abnormal intrusion identification effect of the smart community.

[0032] In some scenarios, the abnormal intrusion monitoring method for smart communities according to an embodiment of the present application can be applied to the abnormal intrusion identification in smart communities and can be used for intrusion detection in areas such as residential areas.

[0033] The following describes in detail a method for monitoring abnormal intrusions in a smart community provided by an embodiment of the present application with reference to specific examples.

[0034] Figure 1 The flowchart of the first abnormal intrusion monitoring method for a smart community provided in the embodiment of the present application is as follows: Figure 1 As shown, the method includes S110 to S130, and S110 to S130 are described in detail below.

[0035] S110: Obtain multiple surveillance videos corresponding to a target person captured by multiple cameras, and determine key frames in the multiple surveillance videos corresponding to the target person, a temporal sequence of the key frames in the multiple surveillance videos, and a time span between the key frames in the multiple surveillance videos. Using an event recognition model, determine, based on the key frames in the multiple surveillance videos, event types and event monitoring indicators corresponding to the key frames in the multiple surveillance videos.

[0036] When conducting intrusion detection in a smart community, multiple surveillance videos corresponding to the target person taken by multiple cameras can be obtained respectively. The target person is any person in the community. This application can identify the surveillance video of the target person and perform event recognition on the surveillance video to identify the activities of the target person.

[0037] After obtaining multiple surveillance videos corresponding to the target person, the key frames of the multiple surveillance videos corresponding to the target person, the temporal order of the key frames of the multiple surveillance videos, and the time span between the key frames of the multiple surveillance videos can be determined using the person event recognition model. The key frames of the multiple surveillance videos are key frames that can represent the activity events of the target person. After obtaining the key frames of the multiple surveillance videos on the timeline, the temporal order of the key frames of the multiple surveillance videos on the timeline can be determined, and the time span between the key frames of the multiple surveillance videos can be determined. The time span is the time interval between the key frames of the multiple surveillance videos.

[0038] Exemplarily, the key frames of the plurality of surveillance videos include key frames corresponding to the target person's residence time, specific actions, and specific behaviors and expressions in a certain area.

[0039] When key frames of multiple surveillance videos corresponding to the target person are obtained, the event recognition model can be used to determine the event types and event monitoring indicators corresponding to the key frames of the multiple surveillance videos based on the key frames of the multiple surveillance videos, and then the activities of the target person can be monitored based on the event types and event monitoring indicators.

[0040] For example, the event type may be an event type recorded in text, and the event monitoring indicator may be a labeled activity suspicion value.

[0041] For example, the event recognition model can be trained by annotating key frames, event types, and event monitoring indicators of surveillance videos.

[0042] S120: Construct a monitoring event graph corresponding to the target person based on the event types and event monitoring indicators corresponding to the key frames of the multiple monitoring videos. The monitoring event graph includes event nodes and edges connecting the event nodes. The attributes of the event nodes include the event type and the event monitoring indicator. The edges connecting the event nodes connect the event nodes in the time sequence of the key frames of the multiple monitoring videos. The attributes of the edges connecting the event nodes include the time span between the event nodes corresponding to the key frames of the monitoring videos.

[0043] After obtaining the event types and event monitoring indicators corresponding to the key frames of multiple surveillance videos, a monitoring event graph corresponding to the target person is constructed based on the event types and event monitoring indicators corresponding to the key frames of multiple surveillance videos. The monitoring event graph represents the overall characteristics of the relationship between the activity events of the target person at different locations in the community.

[0044] It should be noted that the monitoring event graph includes event nodes and edges connecting event nodes. The attributes of event nodes include event types and event monitoring indicators, and the event types and event detection indicators of target personnel can be represented by event nodes; the edges connecting event nodes are used to connect event nodes, and the edges connecting event nodes connect event nodes according to the time sequence of key frames of multiple monitoring videos. The attributes of the edges connecting event nodes include the time span between event nodes corresponding to the key frames of the monitoring video, and the time span between the key frames of the monitoring video can be represented by the edges connecting event nodes.

[0045] S130. Determine the target person's abnormal intrusion probability value based on the monitoring event graph using the abnormal intrusion detection model. When the target person's abnormal intrusion probability value is greater than or equal to a preset abnormal intrusion probability value, issue an abnormal intrusion prompt message corresponding to the target person. The abnormal intrusion detection model is a graph neural network model.

[0046] After obtaining the monitoring event graph, the abnormal intrusion probability value of the target person can be determined according to the monitoring event graph through the abnormal intrusion detection model. The abnormal intrusion probability value represents the abnormal intrusion probability of the target person in the community. The abnormal intrusion detection model is a graph neural network model. The graph neural network model can extract and identify the features of the event nodes and the edges connecting the event nodes of the monitoring event graph, and can identify the relationship between the target person corresponding to the monitoring event graph in the event nodes and the edges connecting the event nodes, that is, it realizes the feature recognition of different events in the monitoring videos of the target person at different locations and the time span between different events, thereby improving the recognition of abnormal intrusion of the target person.

[0047] It should be noted that the abnormal intrusion detection model is a graph neural network model, which can be trained through the labeled monitoring event graph and the abnormal intrusion probability value of personnel.

[0048] After obtaining the abnormal intrusion probability value of the target person, when the abnormal intrusion probability value of the target person is greater than or equal to the preset abnormal intrusion probability value, it means that the probability of the target person being an abnormal intruder is relatively high. At this time, an abnormal intrusion prompt message corresponding to the target person can be issued to prompt the security personnel to handle or review.

[0049] The beneficial effect of the above-mentioned implementation method is that the monitoring event graph includes event nodes and edges connecting event nodes, the attributes of the event nodes include event type and event monitoring indicators, and the attributes of the edges connecting event nodes include the time span between event nodes corresponding to the key frames of the monitoring video. Through the abnormal intrusion detection model, the abnormal intrusion probability value of the target person is determined according to the monitoring event graph, thereby improving the recognition accuracy of the abnormal intrusion probability of the target person.

[0050] Figure 2 The flowchart of the second abnormal intrusion monitoring method for a smart community provided in the embodiment of the present application is as follows: Figure 2 As shown, the above method includes S210 to S220, and S210 to S220 are described in detail below.

[0051] S210. Obtain the regional scenes corresponding to the key frames of multiple surveillance videos, and obtain the maximum movement time, minimum movement time and normal movement time between the regional scenes corresponding to the key frames of multiple surveillance videos that are adjacent in time sequence, and add the maximum movement time, minimum movement time and normal movement time between the regional scenes corresponding to the key frames of multiple surveillance videos in the attributes of the edge connecting the event node.

[0052] When identifying abnormal intrusions, the target person's movement speed can be identified to identify events such as unwarranted wandering or abnormal pausing, thereby improving the accuracy of identifying abnormal intrusions. During identification, the regional scenes corresponding to the key frames of multiple surveillance videos can be obtained, and the maximum movement time, minimum movement time, and normal movement time between the regional scenes corresponding to the key frames of multiple surveillance videos that are adjacent in time sequence can be obtained. The maximum movement time, minimum movement time, and normal movement time between regional scenes can be obtained by security personnel through experiments using maximum walking speed, minimum walking speed, and normal walking speed.

[0053] After obtaining the maximum movement time, minimum movement time, and normal movement time between regional scenes, the maximum movement time, minimum movement time, and normal movement time between regional scenes corresponding to multiple key frames of surveillance videos can be added to the attributes of the edge connecting the event node. Then, the movement speed characteristics of the target person can be further judged based on the maximum movement time, minimum movement time, and normal movement time between regional scenes to identify events such as the target person's unexplained wandering and abnormal stay.

[0054] S220. Determine the abnormal intrusion probability value of the target person according to the monitoring event graph through the abnormal intrusion detection model.

[0055] After adding the maximum movement time, minimum movement time, and normal movement time between the regional scenes corresponding to the key frames of multiple surveillance videos in the attributes of the edges connecting the event nodes, the abnormal intrusion detection model can be used to determine the abnormal intrusion probability value of the target person according to the surveillance event graph, thereby achieving high-precision detection of abnormal intrusion of the target person by combining the movement speed characteristics of the target person.

[0056] The beneficial effect of the above implementation method is that after adding the maximum movement time, minimum movement time and normal movement time between the regional scenes corresponding to the key frames of multiple surveillance videos in the attributes of the edges connecting the event nodes, the abnormal intrusion of the target person can be detected with high precision in combination with the movement speed characteristics of the target person.

[0057] Figure 3 The flowchart of the third abnormal intrusion monitoring method for a smart community provided in the embodiment of the present application is as follows: Figure 3 As shown, the above method includes S310 to S320, and S310 to S320 are described in detail below.

[0058] S310. Determine monitoring event graphs at the moments corresponding to the key frames of multiple monitoring videos, determine the abnormal intrusion probability value of the target person corresponding to each monitoring event graph through the abnormal intrusion detection model, and determine the cumulative value of the abnormal intrusion probability value.

[0059] When detecting abnormal intrusion of target persons, in order to detect abnormal intrusion of target persons in real time, the monitoring event graphs at the moments corresponding to the key frames of multiple monitoring videos can be determined. Different monitoring video key frames are at different moments, that is, the monitoring event graphs correspond to different moments. Then, after updating a key frame of a monitoring video, abnormal intrusion detection can be performed immediately through the abnormal intrusion detection model to realize abnormal intrusion detection of the target person.

[0060] After obtaining the monitoring event graphs at the moments corresponding to the key frames of multiple monitoring videos, the abnormal intrusion probability value of the target person corresponding to each monitoring event graph can be determined through the abnormal intrusion detection model. The abnormal intrusion probability value of the target person corresponding to each monitoring event graph represents the abnormal intrusion probability value at different moments.

[0061] After obtaining the abnormal intrusion probability values ​​at different times, the cumulative value of the abnormal intrusion probability values ​​may be determined. The cumulative value of the abnormal intrusion probability values ​​represents the sum of the abnormal intrusion probabilities of the target user at different times.

[0062] S320: When the cumulative value of the abnormal intrusion probability value of the target person is greater than or equal to the cumulative value of the preset abnormal intrusion probability value, abnormal intrusion prompt information corresponding to the target person is issued.

[0063] After obtaining the cumulative value of the target person's abnormal intrusion probability value, when the cumulative value of the target person's abnormal intrusion probability value is greater than or equal to the cumulative value of the preset abnormal intrusion probability value, it means that the sum of the intrusion probabilities of the target person at different times is too large, and then the abnormal intrusion prompt information corresponding to the target person can be issued to prompt the abnormal intrusion.

[0064] The beneficial effect of the above implementation is that the cumulative value of the abnormal intrusion probability value represents the sum of the abnormal intrusion probabilities of the target user at different times, and the abnormal intrusion of the target user is accurately identified in real time through the cumulative value of the abnormal intrusion probability value.

[0065] Figure 4 The fourth method for monitoring abnormal intrusion in smart communities provided in the embodiment of the present application is a flow chart, as shown in FIG. Figure 4 As shown, the above method includes S410 to S420, and S410 to S420 are described in detail below.

[0066] S410. Obtain the regional scenes to which the key frames of the multiple surveillance videos respectively belong, and obtain the intrusion probability weight value corresponding to each regional scene, and determine the sum of the products of the abnormal intrusion probability values ​​and the intrusion probability weight values ​​corresponding to the regional scenes to which the key frames of the multiple surveillance videos respectively belong as the cumulative value of the abnormal intrusion probability values.

[0067] When detecting abnormal intrusions of a target person, different areas have different importance. Therefore, intrusion detection can be performed based on the importance weights of key frames in surveillance video from different areas. Different areas can include key surveillance areas such as the area around the monitoring room and walls, as well as general areas such as squares and community entrances.

[0068] When performing intrusion detection, the regional scenes to which the key frames of multiple surveillance videos belong can be obtained. The regional scenes are the areas captured by the cameras corresponding to the key frames of multiple surveillance videos, and the intrusion probability weight value corresponding to each regional scene can be obtained. The intrusion probability weight value represents the importance characteristics corresponding to different regional scenes.

[0069] After obtaining the intrusion probability weight value corresponding to each regional scene, the sum of the products of the abnormal intrusion probability value and the intrusion probability weight value corresponding to the regional scenes to which the key frames of multiple surveillance videos belong can be determined. The abnormal intrusion probability values ​​corresponding to multiple regional scenes are summed according to the intrusion probability weight value, and the sum is used as the cumulative value of the abnormal intrusion probability value, thereby adjusting the abnormal intrusion detection weight of key areas.

[0070] S420: When the cumulative value of the abnormal intrusion probability value of the target person is greater than or equal to the cumulative value of the preset abnormal intrusion probability value, abnormal intrusion prompt information corresponding to the target person is issued.

[0071] After obtaining the cumulative value of the target person's abnormal intrusion probability value, when the cumulative value of the target person's abnormal intrusion probability value is greater than or equal to the cumulative value of the preset abnormal intrusion probability value, abnormal intrusion prompt information corresponding to the target person can be issued.

[0072] The beneficial effect of the above implementation method is that the abnormal intrusion probability values ​​corresponding to multiple area scenes are summed according to the intrusion probability weight value, and used as the cumulative value of the abnormal intrusion probability value, thereby adjusting the abnormal intrusion detection weights of key areas and improving the accuracy of abnormal intrusion detection of target personnel.

[0073] Figure 5 The flowchart of the fifth abnormal intrusion monitoring method for a smart community provided in the embodiment of the present application is as follows: Figure 5 As shown, the above method includes S510 to S520, and S510 to S520 are described in detail below.

[0074] S510: Obtain regional scenes to which key frames of multiple surveillance videos respectively belong, obtain a key frame quantity threshold and a time span threshold for surveillance videos corresponding to each regional scene, and determine the number of key frames of surveillance videos in each regional scene.

[0075] When monitoring different areas, different areas may have different requirements on the number of key frames and time span when the monitoring importance is different, thereby ensuring the monitoring accuracy of areas of different importance.

[0076] When monitoring areas of different importance, the regional scenes to which the key frames of multiple surveillance videos belong can be obtained, and the key frame number threshold and time span threshold of the surveillance video corresponding to each regional scene can be obtained. The key frame number threshold and time span threshold of the surveillance video corresponding to each regional scene can be preset.

[0077] After obtaining the key frame number threshold of the surveillance video corresponding to each regional scene, the number of key frames of the surveillance video in each regional scene can be determined, and then abnormal intrusion identification can be performed based on the number of key frames of the surveillance video in each regional scene.

[0078] S520. When the number of key frames of the surveillance video corresponding to the regional scene is less than the key frame number threshold of the surveillance video, or the time span of the key frames of the surveillance video corresponding to the regional scene is less than the preset time span threshold, delete the event node corresponding to the key frame of the surveillance video corresponding to the regional scene in the monitoring event graph.

[0079] When identifying abnormal intrusions, when the number of key frames of the surveillance video corresponding to the regional scene is less than the key frame number threshold of the surveillance video, it means that the number of key frames of the surveillance video in the regional scene does not meet the basic requirements for monitoring abnormal intrusions. At this time, in order to avoid introducing noise, the event nodes corresponding to the key frames of the surveillance video corresponding to the regional scene can be deleted in the monitoring event graph to avoid the impact of abnormal intrusions caused by too few key frames of the surveillance video corresponding to the regional scene.

[0080] Similarly, when the time span of the key frame of the surveillance video corresponding to the regional scene is less than the preset time span threshold, it means that the time span of the key frame of the surveillance video corresponding to the regional scene does not meet the basic requirements of intrusion detection, and then the event node corresponding to the key frame of the surveillance video corresponding to the regional scene can be deleted in the monitoring event graph.

[0081] The beneficial effect of the above-mentioned implementation method is to obtain the key frame number threshold and time span threshold of the surveillance video corresponding to each regional scene, and filter the key frames of the surveillance video corresponding to each regional scene through the key frame number threshold and time span threshold of the surveillance video corresponding to each regional scene, thereby avoiding noise in abnormal intrusion detection and improving the accuracy of abnormal intrusion detection of target personnel.

[0082] In some implementations, the above method includes S610 to S620, and S610 to S620 are described in detail below.

[0083] S610: Obtain time periods to which key frames of multiple surveillance videos belong, and obtain time period adjustment factors corresponding to the time periods to which key frames of the multiple surveillance videos belong.

[0084] When detecting abnormal intrusions, the flow of people in different time periods may be different, and the probability of possible abnormal intrusions may also be different. Therefore, the effect of abnormal intrusion detection can be adjusted according to the characteristics of the time period.

[0085] When adjusting the intrusion detection effect, the time periods to which the key frames of multiple surveillance videos belong can be obtained, and the time period adjustment factors corresponding to the time periods to which the key frames of multiple surveillance videos belong can be obtained. The time period adjustment factors represent the time adjustment characteristics of the time periods to which the key frames of multiple surveillance videos belong.

[0086] For example, the time period adjustment factor corresponding to the time period to which the key frame of the surveillance video belongs may be preset according to different time periods.

[0087] S620: Determine the product of a key frame number threshold and a time period adjustment factor for the surveillance video corresponding to each regional scene, and adjust the key frame number threshold. Determine the product of a time span threshold and a time period adjustment factor for the surveillance video corresponding to each regional scene, and adjust the time period adjustment factor.

[0088] After obtaining the time period adjustment factors corresponding to the time periods to which the key frames of multiple surveillance videos belong, the product of the key frame number threshold of the surveillance video corresponding to each regional scene and the time period adjustment factor can be determined to adjust the key frame number threshold, and then the key frame number threshold can be adjusted according to the pedestrian flow characteristics of the time period to which it belongs.

[0089] Similarly, the product of the time span threshold of the surveillance video corresponding to each regional scene and the time period adjustment factor can be determined to adjust the time period adjustment factor, and then the time span threshold can be adjusted according to the pedestrian flow characteristics of the time period.

[0090] The beneficial effect of the above implementation method is that the key frame number threshold is adjusted according to the pedestrian flow characteristics of the time period to which it belongs, and the time span threshold is adjusted according to the pedestrian flow characteristics of the time period to which it belongs, thereby improving the accuracy of abnormal intrusion detection of target personnel.

[0091] Figure 6 The sixth method for monitoring abnormal intrusion in smart communities provided in the embodiment of the present application is a flow chart, as shown in FIG. Figure 6 As shown, the above method includes S710 to S720, and S710 to S720 are described in detail below.

[0092] S710: Determine multiple associated persons of the target person, and construct monitoring event graphs corresponding to the multiple associated persons.

[0093] When performing abnormal intrusion detection, the abnormal intrusion group can be detected to improve the detection effect of abnormal intrusion, while avoiding the insufficient accuracy of abnormal intrusion detection on a single abnormal intruder.

[0094] When performing group abnormal intrusion detection, multiple associated persons of the target person can be determined, and monitoring event graphs corresponding to the multiple associated persons can be constructed. The monitoring event graphs corresponding to the multiple associated persons are monitoring event graphs corresponding to the multiple associated persons related to the target person.

[0095] S720. Determine, using the abnormal intrusion detection model, a group abnormal intrusion probability value corresponding to the target person and the multiple associated persons based on the monitoring event graphs corresponding to the target person and the multiple associated persons. When the group abnormal intrusion probability value corresponding to the target person and the multiple associated persons is greater than or equal to a preset group abnormal intrusion probability value, issue a group abnormal intrusion prompt message corresponding to the target person and the multiple associated persons.

[0096] When performing group abnormal intrusion probability value detection, the abnormal intrusion detection model can be used to determine the group abnormal intrusion probability value corresponding to the target person and multiple related persons based on the monitoring event graphs corresponding to the target person and multiple related persons respectively. The group abnormal intrusion probability value represents the probability that the target person and multiple related persons as a whole are judged to be abnormal intrusions.

[0097] Exemplarily, the group abnormal intrusion detection model can be obtained by training the monitoring event graph corresponding to the marked target person and multiple related persons, and the group intrusion probability value data.

[0098] After obtaining the group intrusion probability value, when the group abnormal intrusion probability value corresponding to the target person and multiple related persons is greater than or equal to the preset group abnormal intrusion probability value, it means that the group intrusion probability of the target person and multiple related persons is too high. At this time, a group abnormal intrusion prompt information corresponding to the target person and multiple related persons can be issued.

[0099] The beneficial effect of the above-mentioned implementation method is that, through the monitoring event graphs corresponding to the target person and multiple related persons respectively, the abnormal intrusion probability values ​​corresponding to the target person and multiple related persons respectively are identified, and the group intrusion probability values ​​corresponding to the target person and multiple related persons can be accurately identified, thereby improving the monitoring accuracy of group abnormal intrusion.

[0100] In some implementations, in the above S710, determining multiple associated persons of the target person includes S711 and S712.

[0101] S711. Construct a monitoring event graph corresponding to multiple personnel respectively, determine the event nodes corresponding to the same area scene in the monitoring event graph of each personnel and the target personnel respectively; determine the similarity of the event nodes corresponding to the same area scene of each personnel and the target personnel respectively, and determine the sum of the similarities of all event nodes of each personnel and the target personnel as the personnel association degree.

[0102] When identifying the associated persons of the target person, a monitoring event graph corresponding to multiple persons can be constructed to determine the event nodes corresponding to the same area scenes in the monitoring event graph of each person and the target person. The event nodes corresponding to the same area scenes of the target person and the associated persons indicate that the target person and the associated persons appear in the same area at the same time, so there may be a possibility of abnormal group intrusion.

[0103] When identifying the associated persons of the target person, the similarity of the event nodes corresponding to each person and the target person in the same area scene can be determined, and the sum of the similarities of all event nodes of each person and the target person can be determined as the person association degree. The person association degree characterizes the possibility of each person and the target person as a group abnormal intrusion in different areas.

[0104] S712: When the person association degree of the first person is greater than or equal to the preset person association degree, the first person is used as an associated person of the target person.

[0105] After obtaining the personnel association degree of each person, when the personnel association degree of the first person is greater than or equal to the preset personnel association degree, the first person is taken as the associated person of the target person, and the group abnormal intrusion probability of the associated persons can be identified.

[0106] The beneficial effect of the above-mentioned implementation method is to determine the similarity of the event nodes corresponding to each person and the target person in the same area scene, and to determine the sum of the similarities of all event nodes of each person and the target person as the person association degree. When the person association degree of the first person is greater than or equal to the preset person association degree, the first person is regarded as the associated person of the target person, and the associated persons can be used to identify the probability of abnormal intrusion of the group.

[0107] In some implementations, the above method includes: obtaining a similarity adjustment factor corresponding to each regional scene, and determining the product of the similarity of the event nodes corresponding to each person and the target person in the same regional scene and the similarity adjustment factor as the person association.

[0108] When determining the personnel association, due to the different characteristics of pedestrian flow in different areas, for example, people are more likely to gather at the entrance of the community, while people are not easy to gather in the area around the community wall. Therefore, different similarity adjustment factors can be set for different areas, and the similarity adjustment factor corresponding to each area scene can be obtained. The product of the similarity of the event nodes corresponding to each person and the target person in the same area scene and the similarity adjustment factor is determined as the personnel association, so that the personnel association can be adjusted according to the characteristics of the area, thereby improving the accuracy of identifying the associated persons of the target person and improving the accuracy of abnormal group intrusion.

[0109] The beneficial effect of the above implementation method is that the personnel association degree can be adjusted according to regional characteristics, thereby improving the accuracy of identifying the target person's associated persons and improving the accuracy of abnormal group intrusion.

[0110] An embodiment of the present application also provides an abnormal intrusion monitoring system for a smart community, comprising a unit for executing any of the methods described above.

[0111] Figure 7 A logical structure diagram of an abnormal intrusion monitoring system for a smart community provided in an embodiment of the present application is shown as follows: Figure 7 As shown, the system 1 of this embodiment includes a processing unit 11, a storage unit 12, and a transceiver unit 13. The processing unit 11 is used to process data, the storage unit 12 is used to store data, and the transceiver unit 13 is used to send and receive data. The processing unit 11, the storage unit 12, and the transceiver unit 13 cooperate with each other to implement the above method. The beneficial effects of the embodiment of the present application have been described in the above method and will not be repeated here.

[0112] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of this application. Their specific functions and technical effects can be found in the method embodiment section and will not be repeated here.

[0113] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0114] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the process of the above-mentioned method embodiment by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can at least include: any entity or device capable of carrying computer program code to the camera / terminal device, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium. For example, a USB flash drive, mobile hard drive, magnetic disk, or optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals or telecommunication signals.

[0115] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0116] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0117] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0118] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0119] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application 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. 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 various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A method for monitoring abnormal intrusion in smart communities, characterized in that: The method comprises: Acquire multiple surveillance videos corresponding to the target person captured by multiple cameras, and determine the key frames of the multiple surveillance videos corresponding to the target person, the time sequence of the key frames of the multiple surveillance videos, and the time span between the key frames of the multiple surveillance videos; determine the event types and event monitoring indicators corresponding to the key frames of the multiple surveillance videos based on the event recognition model; Constructing a monitoring event graph corresponding to a target person based on the event types and event monitoring indicators corresponding to the key frames of the multiple monitoring videos; wherein the monitoring event graph includes event nodes and edges connecting the event nodes, the attributes of the event nodes include the event type and the event monitoring indicator, the edges connecting the event nodes connect the event nodes in the time sequence of the key frames of the multiple monitoring videos, and the attributes of the edges connecting the event nodes include the time span between the event nodes corresponding to the key frames of the monitoring videos; Through the abnormal intrusion detection model, according to the monitoring event graph, the abnormal intrusion probability value of the target person is determined. When the abnormal intrusion probability value of the target person is greater than or equal to the preset abnormal intrusion probability value, the abnormal intrusion prompt information corresponding to the target person is issued; wherein, the abnormal intrusion detection model is a graph neural network model; The method comprises: Obtaining the regional scenes to which the key frames of the multiple surveillance videos respectively belong, and obtaining the key frame quantity threshold and time span threshold of the surveillance video corresponding to each regional scene; determining the number of key frames of the surveillance video in each regional scene; When the number of key frames of the surveillance video corresponding to the regional scene is less than the key frame number threshold of the surveillance video, or when the time span of the key frames of the surveillance video corresponding to the regional scene is less than the preset time span threshold, the event node corresponding to the key frame of the surveillance video corresponding to the regional scene is deleted in the monitoring event graph; The method comprises: Obtaining time periods to which key frames of multiple surveillance videos belong, and obtaining time period adjustment factors corresponding to the time periods to which key frames of the multiple surveillance videos belong; Determine the product of the key frame number threshold and the time period adjustment factor of the surveillance video corresponding to each regional scene to adjust the key frame number threshold; determine the product of the time span threshold and the time period adjustment factor of the surveillance video corresponding to each regional scene to adjust the time period adjustment factor.

2. The method according to claim 1, wherein The method comprises: Obtain the regional scenes corresponding to the key frames of multiple surveillance videos, and obtain the maximum movement time, minimum movement time, and normal movement time between the regional scenes corresponding to the key frames of multiple surveillance videos that are adjacent in time sequence, and add the maximum movement time, minimum movement time, and normal movement time between the regional scenes corresponding to the key frames of multiple surveillance videos to the attributes of the edge connecting the event node; Through the abnormal intrusion detection model, the abnormal intrusion probability value of the target person is determined according to the monitoring event graph.

3. The method according to claim 2, wherein The method comprises: Determine the monitoring event graphs at the moments corresponding to the key frames of multiple monitoring videos, determine the abnormal intrusion probability value of the target person corresponding to each monitoring event graph through the abnormal intrusion detection model, and determine the cumulative value of the abnormal intrusion probability value; When the cumulative value of the abnormal intrusion probability value of the target person is greater than or equal to the cumulative value of the preset abnormal intrusion probability value, abnormal intrusion prompt information corresponding to the target person is issued.

4. The method according to claim 3, wherein The method comprises: Obtaining the regional scenes to which the key frames of the multiple surveillance videos respectively belong, and obtaining the intrusion probability weight value corresponding to each regional scene, and determining the sum of the products of the abnormal intrusion probability values ​​and the intrusion probability weight values ​​corresponding to the regional scenes to which the key frames of the multiple surveillance videos respectively belong as the cumulative value of the abnormal intrusion probability values; When the cumulative value of the abnormal intrusion probability value of the target person is greater than or equal to the cumulative value of the preset abnormal intrusion probability value, abnormal intrusion prompt information corresponding to the target person is issued.

5. The method according to claim 4, wherein The method comprises: Identify multiple associated persons of the target person and construct monitoring event graphs corresponding to the multiple associated persons; Through the abnormal intrusion detection model, according to the monitoring event graphs corresponding to the target person and multiple related persons respectively, the group abnormal intrusion probability value corresponding to the target person and multiple related persons is determined; when the group abnormal intrusion probability value corresponding to the target person and multiple related persons is greater than or equal to the preset group abnormal intrusion probability value, the group abnormal intrusion prompt information corresponding to the target person and multiple related persons is issued.

6. The method according to claim 5, wherein Identify multiple associates of the target person, including: Construct a monitoring event graph corresponding to multiple personnel, determine the event nodes corresponding to the same regional scenes in the monitoring event graph of each personnel and the target personnel; determine the similarity of the event nodes corresponding to each personnel and the target personnel in the same regional scenes, and determine the sum of the similarities of all event nodes of each personnel and the target personnel as the personnel association degree; When the person association degree of the first person is greater than or equal to the preset person association degree, the first person is taken as an associated person of the target person.

7. The method according to claim 6, wherein The method comprises: The similarity adjustment factor corresponding to each regional scene is obtained, and the product of the similarity of the event nodes corresponding to each person and the target person in the same regional scene and the similarity adjustment factor is determined as the person association degree.

8. An abnormal intrusion monitoring system for a smart community, characterized in that: Comprising means for performing the method according to any one of claims 1 to 7.

Images