A cross-camera video recommendation method and device

Abstract

The application provides a cross-camera video recommendation method and device, electronic equipment and computer program product, and relates to the technical field of video monitoring. The method comprises the following steps: constructing a pedestrian route network according to cross-camera pedestrian re-identification records of all monitoring device nodes in a monitoring platform; calculating the propagation efficiency value of each monitoring device node by using a pre-constructed node propagation efficiency algorithm model based on the pedestrian route network; determining a monitoring device node whose propagation efficiency value meets a preset first condition as a key monitoring device node of the monitoring platform; and performing a hotspot area real-time monitoring strategy on the key monitoring device node. The embodiments of the application can automatically identify the key areas of the current monitoring platform through an algorithm, so as to dynamically recommend real-time hotspot videos for these key areas, and thus effectively improve the real-time performance and accuracy of hotspot video selection in a cross-camera monitoring scene.

Description

Technical Field

[0001] This application relates to the field of video surveillance technology, specifically to a method, apparatus, electronic device, and computer program product for video recommendation across cameras. Background Technology

[0002] Video surveillance technology, as a crucial means of maintaining public safety, plays a vital role in combating illegal and criminal activities, enhancing the emergency response capabilities of security personnel, and promoting the smooth operation and management of cities. With the continuous development of technology, visual (pedestrian) target tracking is an important research direction in computer vision. In recent years, with the advancement of deep learning technology, target tracking technology has made groundbreaking progress. The detection range of a single camera is limited; when tracking a target, if the target leaves the camera's field of view, the target will be lost. Through cross-camera collaborative relay tracking of pedestrians, assessing and effectively focusing on important areas significantly enhances the real-time hotspot monitoring capabilities of surveillance platforms.

[0003] Currently, cross-camera monitoring platforms primarily rely on manually selecting multiple cameras to achieve basic monitoring of fixed areas. However, when the monitoring platform covers a large area and has a large number of connected cameras, effective monitoring of real-time hotspots cannot be guaranteed, requiring manual switching by monitoring personnel. In multi-camera monitoring platforms, real-time hotspots refer to important streets, densely populated areas, or transportation hubs requiring attention. Because pedestrian and vehicular traffic changes dynamically in real time, the traditional method of manually selecting monitoring equipment only allows for a rough selection of key areas based on intuitive manual assessment, resulting in low real-time performance and accuracy in video selection during real-time hotspot monitoring scenarios. Summary of the Invention

[0004] This application provides a method, apparatus, electronic device, and computer program product for recommending videos across cameras, in order to solve the above-mentioned technical problems and improve the real-time performance and accuracy of hotspot video selection in cross-camera video surveillance scenarios.

[0005] In a first aspect, embodiments of this application provide a cross-camera video recommendation method, including:

[0006] Based on the cross-camera pedestrian re-identification records of all monitoring device nodes in the monitoring platform, a pedestrian route network is constructed.

[0007] Based on the pedestrian route network, the propagation efficiency value of each monitoring device node is calculated using a pre-built node propagation efficiency algorithm model.

[0008] The monitoring device nodes whose propagation efficiency values ​​meet the preset first condition are identified as the key monitoring device nodes of the monitoring platform.

[0009] A real-time monitoring strategy for hotspot areas is implemented on the key monitoring equipment nodes.

[0010] In one embodiment, the cross-camera video recommendation method further includes:

[0011] When a target monitoring device node triggers a sudden event monitoring request, the correlation between each monitoring device node and the target monitoring device node is calculated based on the pedestrian route network using a pre-built node correlation algorithm model.

[0012] The monitoring device node whose relevance meets the preset second condition is determined as the associated monitoring device node of the target monitoring device node;

[0013] Implement emergency monitoring strategies for sudden events on the associated monitoring device nodes.

[0014] In one embodiment, the execution of a real-time hotspot monitoring strategy on the key monitoring device nodes includes:

[0015] The monitoring cycle and video recording duration of the corresponding key monitoring equipment nodes are determined based on the magnitude of each propagation efficiency value.

[0016] Based on a defined monitoring cycle, dynamic video feeds are pushed to the areas corresponding to each key monitoring device node, and video recordings and storage are performed on the areas corresponding to each key monitoring device node based on a defined video recording duration.

[0017] In one embodiment, executing an emergency monitoring strategy for sudden events on the associated monitoring device node includes:

[0018] The monitoring cycle and video recording duration of the corresponding associated monitoring device nodes are determined based on the magnitude of each relevant value.

[0019] Based on a defined monitoring cycle, dynamic video feeds are pushed to the areas corresponding to each associated monitoring device node, and video recordings and storage are performed on the areas corresponding to each associated monitoring device node based on a defined video recording duration.

[0020] In one embodiment, determining the monitoring device node whose propagation efficiency value meets a preset first condition as the key monitoring device node of the monitoring platform includes:

[0021] All monitoring device nodes are sorted in descending order of propagation efficiency values, and the top-ranked nodes are identified as key monitoring device nodes for the monitoring platform; or,

[0022] Monitoring device nodes whose propagation efficiency values ​​are greater than a preset first threshold are identified as key monitoring device nodes of the monitoring platform.

[0023] In one embodiment, determining the monitoring device node whose relevance meets the preset second condition as the associated monitoring device node of the target monitoring device node includes:

[0024] All monitoring device nodes are sorted in descending order of relevance, and the top-ranked nodes are identified as associated monitoring device nodes of the target monitoring device node; or,

[0025] Monitoring device nodes with a relevance greater than a preset second threshold are identified as associated monitoring device nodes of the target monitoring device node.

[0026] In one embodiment, the node propagation efficiency algorithm model is the PageRank algorithm model, and the node correlation algorithm model is the PersonalRank algorithm model.

[0027] Secondly, embodiments of this application provide a cross-camera video recommendation device, comprising:

[0028] The network construction module is used to construct a pedestrian route network based on the cross-camera pedestrian re-identification records of all monitoring device nodes in the monitoring platform;

[0029] The propagation efficiency calculation module is used to calculate the propagation efficiency value of each monitoring device node based on the pedestrian route network and using a pre-built node propagation efficiency algorithm model.

[0030] The critical node determination module is used to determine the monitoring device nodes whose propagation efficiency values ​​meet a preset first condition as the critical monitoring device nodes of the monitoring platform.

[0031] The hotspot area monitoring module is used to execute real-time hotspot area monitoring strategies on the key monitoring device nodes.

[0032] Thirdly, embodiments of this application provide an electronic device, including a processor and a memory storing a computer program, wherein the processor executes the program to implement the steps of the cross-camera video recommendation method described in the first aspect.

[0033] Fourthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the cross-camera video recommendation method described in the first aspect.

[0034] The cross-camera video recommendation method, apparatus, electronic device, and computer program product provided in this application calculate the propagation efficiency value of each monitoring device node on the monitoring platform using a pre-built algorithm model. Based on the propagation efficiency value, key nodes within the monitoring area are selected, and real-time monitoring of hotspot areas is implemented based on the selected key nodes, eliminating the need for manual selection of monitoring devices through subjective human judgment. This application embodiment can automatically identify key areas of the current monitoring platform through algorithms, thereby enabling real-time dynamic recommendation of hotspot videos for these key areas, effectively improving the real-time performance and accuracy of hotspot video selection in cross-camera monitoring scenarios. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a flowchart illustrating the cross-camera video recommendation method provided in an embodiment of this application;

[0037] Figure 2 This is a schematic diagram of the cross-camera video recommendation device provided in an embodiment of this application;

[0038] Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0040] Figure 1 This is a flowchart illustrating a cross-camera video recommendation method. (Refer to...) Figure 1 This application provides a cross-camera video recommendation method, which may include the following steps:

[0041] S1. Construct a pedestrian route network based on the cross-camera pedestrian re-identification records of all monitoring device nodes in the monitoring platform;

[0042] S2. Based on the pedestrian route network, calculate the propagation efficiency value of each monitoring device node using a pre-built node propagation efficiency algorithm model;

[0043] S3. The monitoring device nodes whose propagation efficiency values ​​meet the preset first condition are identified as the key monitoring device nodes of the monitoring platform.

[0044] S4. Implement a real-time monitoring strategy for hotspot areas on the key monitoring equipment nodes.

[0045] In this embodiment of the application, pedestrian re-identification is first performed based on the image data of the monitoring system, pedestrian information across cameras is recorded, and a pedestrian route network is constructed based on the recorded pedestrian re-identification information. The pedestrian route network may include multiple monitoring device nodes, the route relationships between monitoring device nodes, and the corresponding pedestrian re-identification probabilities.

[0046] Then, based on the pedestrian route network, the propagation efficiency value of each monitoring device node is calculated using a pre-built node propagation efficiency algorithm model, such as the PageRank algorithm model.

[0047] Then, based on the calculated propagation efficiency values ​​of each monitoring device node, key monitoring device nodes are selected from all monitoring device nodes. The areas corresponding to these key monitoring device nodes are designated as key monitoring areas. It can be understood that the higher the propagation efficiency value, the more attention the corresponding monitoring device node needs to receive. Furthermore, there can be one or more identified key monitoring device nodes, but generally multiple. It can be understood that the preset first condition can be that the propagation efficiency value is greater than a certain preset threshold (the specific threshold can be set according to actual needs), or that the ranking of the propagation efficiency value is within a preset ranking (the specific ranking threshold can be set according to actual needs).

[0048] After identifying key monitoring device nodes, the areas corresponding to these nodes can be prioritized for monitoring based on preset hotspot area real-time monitoring strategies. For example, the monitoring videos of these key monitoring device nodes can be displayed in a hotspot display area on the monitoring platform.

[0049] The cross-camera video recommendation method provided in this application can automatically identify key areas of the current monitoring platform through algorithms, thereby enabling real-time dynamic recommendation of hotspot videos for these key areas, and effectively improving the real-time performance and accuracy of hotspot video selection in cross-camera monitoring scenarios.

[0050] In one embodiment, the cross-camera video recommendation method may further include the following steps:

[0051] S5. When the target monitoring device node triggers a sudden event monitoring request, based on the pedestrian route network, the correlation between each monitoring device node and the target monitoring device node is calculated using a pre-built node correlation algorithm model; the node correlation algorithm model is the PersonalRank algorithm model.

[0052] S6. The monitoring device node whose relevance meets the preset second condition is determined as the associated monitoring device node of the target monitoring device node;

[0053] S7. Execute the emergency monitoring strategy for the associated monitoring device nodes.

[0054] In this embodiment, in addition to routine monitoring of hotspot areas, the system also includes recommending video feeds associated with devices during emergencies. Specifically, when a target monitoring device node triggers an emergency monitoring request (the target monitoring device node can be automatically identified by image recognition technology or manually selected based on the actual situation), the relevance between all monitoring device nodes in the monitoring platform and this target monitoring device node is calculated. Then, corresponding associated monitoring device nodes are selected based on the relevance, thereby enabling focused monitoring of both the target monitoring device node and its associated monitoring device nodes. Similar to real-time monitoring of hotspot areas, the emergency monitoring strategy can be configured according to actual needs.

[0055] It is understood that the preset second condition may be that the relevance is greater than a certain preset relevance threshold (the specific threshold can be set according to actual needs), or that the relevance ranking is within a preset ranking (the specific ranking threshold can be set according to actual needs).

[0056] The cross-camera video recommendation method provided in this application, when a monitoring device node triggers a sudden event, determines the device objects that need to be focused on by calculating the correlation between each monitoring device node and the target monitoring device node. This enables the automatic identification of nodes associated with the sudden event node, allowing for emergency monitoring of these related areas, shortening the operation time for management personnel, increasing the ability to focus on key monitoring areas, and improving the degree of automation control of the monitoring platform.

[0057] In one embodiment, step S4 may include:

[0058] S41. Determine the monitoring cycle and video recording duration of the corresponding key monitoring equipment nodes based on the magnitude of each propagation efficiency value.

[0059] S42. Based on a defined monitoring cycle, dynamically push video to the areas corresponding to each key monitoring device node, and record and store video to the areas corresponding to each key monitoring device node based on a defined video recording duration.

[0060] It should be noted that after identifying key monitoring device nodes, the areas corresponding to these nodes can be monitored in a focused manner according to a preset hotspot area real-time monitoring strategy. In this embodiment, the monitoring cycle and video recording duration of a key monitoring device node can be determined based on its propagation efficiency value. The monitoring cycle refers to the frequency at which the node's video is pushed to the key focus modules of the monitoring platform; the video recording duration refers to the retention period for recording and storing the node's video, such as storing and retaining monitoring videos from the most recent 24 hours. The specific correspondence between the propagation efficiency value, monitoring cycle, and video recording duration can be preset according to actual conditions. It can be understood that the higher the propagation efficiency value, the shorter the monitoring cycle (higher push frequency) and the longer the video recording duration for that node.

[0061] The cross-camera video recommendation method provided in this application determines the corresponding monitoring cycle and video recording duration based on the different propagation efficiency values ​​of key monitoring device nodes. This allows for adaptive adjustment of attention according to the hotspot status of monitoring device nodes, further improving the real-time performance and accuracy of hotspot video selection.

[0062] In one embodiment, step S7 may include:

[0063] S71. Determine the monitoring cycle and video recording duration of the corresponding associated monitoring device nodes based on the magnitude of each relevant value;

[0064] S72. Based on a defined monitoring cycle, dynamically push video to the areas corresponding to each associated monitoring device node, and record and store video to the areas corresponding to each associated monitoring device node based on a defined video recording duration.

[0065] It should be noted that for a selected (or monitored) target monitoring device node, the correlation between other monitoring device nodes and the target monitoring device node is calculated to identify the associated monitoring device nodes with stronger correlations. When a sudden event occurs at the target monitoring device node requiring close monitoring, the same principle applies to these associated monitoring device nodes; the corresponding monitoring cycle and video recording duration can be determined based on the different correlations between each key monitoring device node and the target monitoring device node. Specific correspondences and monitoring strategies can be set according to specific circumstances.

[0066] The cross-camera video recommendation method provided in this application determines the corresponding monitoring cycle and video recording duration based on the different correlations between associated monitoring device nodes and target monitoring device nodes. This allows for adaptive adjustment of attention based on the degree of correlation between associated and target monitoring device nodes, further improving the real-time performance and accuracy of selecting related videos for areas affected by emergencies.

[0067] In one embodiment, step S3 may include:

[0068] All monitoring device nodes are sorted in descending order of propagation efficiency values, and the top-ranked nodes are identified as key monitoring device nodes for the monitoring platform; or,

[0069] Monitoring device nodes whose propagation efficiency values ​​are greater than a preset first threshold are identified as key monitoring device nodes of the monitoring platform.

[0070] It should be noted that there are two ways to determine key monitoring device nodes: one is to select the top few nodes after sorting them by propagation efficiency value from largest to smallest, for example, selecting the three monitoring device nodes with the largest propagation efficiency value as key monitoring device nodes; the other is to select the monitoring device nodes whose propagation efficiency value is greater than a preset first threshold, for example, if the first threshold is set to 0.15, then the monitoring device nodes with a propagation efficiency value greater than 0.15 are selected as key monitoring device nodes.

[0071] The cross-camera video recommendation method provided in this application determines whether a node is included in the key monitoring target by using the propagation efficiency value of the monitoring device node. This enables more accurate and real-time selection of hot videos for dynamic push in cross-camera video monitoring scenarios.

[0072] In one embodiment, step S6 may include:

[0073] All monitoring device nodes are sorted in descending order of relevance, and the top-ranked nodes are identified as associated monitoring device nodes of the target monitoring device node; or,

[0074] Monitoring device nodes with a relevance greater than a preset second threshold are identified as associated monitoring device nodes of the target monitoring device node.

[0075] It should be noted that there are two ways to determine the associated monitoring device nodes: one is to select the top-ranked nodes after sorting them by their relevance to the target monitoring device node, for example, selecting the three monitoring device nodes with the highest relevance as associated monitoring device nodes; the other is to select the monitoring device nodes with a relevance greater than a preset threshold, for example, if the second threshold is set to 0.1, then the monitoring device nodes with a relevance (represented by the propagation efficiency value) greater than 0.1 are selected as key monitoring device nodes.

[0076] The cross-camera video recommendation method provided in this application determines whether a node should be included as a key monitoring object in case of an emergency by determining the correlation between the monitoring device node and the target monitoring device node. In cross-camera video monitoring scenarios, it can more accurately and in real time select and dynamically push relevant area videos when an emergency occurs in the monitoring area.

[0077] Based on the above solution, and to facilitate a better understanding of the cross-camera video recommendation method provided in this application embodiment, the following detailed explanation is provided:

[0078] It should be noted that current cross-camera monitoring methods on surveillance platforms primarily rely on manually selecting multiple cameras to achieve basic monitoring of fixed areas. When the monitoring platform covers a large area and has a large number of connected cameras, effective monitoring of real-time hotspots cannot be guaranteed, requiring manual switching by monitoring personnel. Furthermore, when a sudden incident occurs within the monitoring area of ​​a camera, and it is necessary to predict the affected monitoring areas nearby, a high level of expertise is required from the monitoring platform administrators. In conclusion, recommending video feeds from monitoring devices for real-time hotspots and from cameras with related characteristics plays a crucial role in practical applications.

[0079] Existing technologies and research in cross-camera video surveillance primarily focus on pedestrian re-identification. To address the issues of subsequent real-time hotspot area recommendation and device-related video recommendation, this application provides a cross-camera video recommendation method that optimizes the monitoring process. By extracting cross-camera re-identification information when pedestrians pass by and analyzing it using the PageRank algorithm, dynamic real-time hotspot video recommendation is achieved on the monitoring platform. For device-related recommendation in the event of an emergency, the PersonalRank algorithm model is used to determine the relevance degree with a specific monitoring device node based on weights, thereby ranking the relevance of monitoring device nodes and providing targeted recommendation of related videos. Implementing this application embodiment will have significant implications for maintaining public safety and promoting the smooth operation and management of cities.

[0080] In multi-camera surveillance platforms, real-time hotspots refer to important streets, densely populated areas, or transportation hubs that require monitoring. Because pedestrian and vehicle traffic changes dynamically in real time, traditional methods of manually selecting surveillance equipment lack real-time accuracy and can only roughly select key areas based on intuitive assessment. This application addresses the dynamic nature of real-time hotspot recommendations by employing the PageRank model, a static algorithm that calculates cross-camera propagation efficiency with relatively low computational cost. This model calculates weights for each surveillance device node within the platform, prioritizing devices with higher weights.

[0081] For monitoring platforms, the propagation efficiency value of each monitoring device node is determined by the PageRank algorithm model, thereby identifying key monitoring device nodes. Based on the propagation efficiency value of key monitoring device nodes, dynamic recommendations of hot devices are achieved.

[0082] When a sudden emergency occurs within the area monitored by a camera, and on-site personnel need to simultaneously observe whether nearby monitored areas are affected, the PersonalRank model is used to recommend videos based on the correlation between monitoring device nodes. Specifically, for cameras that detect sudden events (such as panic attacks), video data from the monitoring platform is extracted, and the historical paths of each pedestrian crossing the camera over a period of time are calculated to obtain the set of devices with the highest correlation, which is then recommended to the monitoring platform.

[0083] The cross-camera video recommendation method provided in this application includes the following steps:

[0084] 1) The image acquisition module provides the system with monitoring image data, generating acquired data. Based on the Person ReID algorithm, it records pedestrian information across devices;

[0085] 2) Based on the pedestrian re-identification records of the multiple monitoring device nodes, a pedestrian route network is constructed, wherein the pedestrian route network includes the multiple monitoring device nodes, the route relationships between the monitoring device nodes, and the corresponding pedestrian re-identification probabilities;

[0086] 3) Based on the constructed pedestrian route network, calculate the propagation efficiency value of each monitoring device node, and conduct real-time hotspot area assessment for the key monitoring device nodes;

[0087] 4) In the event of an emergency, based on the recommendation algorithm of the surveillance equipment video with the correlation of the surveillance equipment nodes and the pedestrian route network, the platform recommends the video with the highest relevance.

[0088] I. The specific steps for recommending device videos in real-time hotspot areas are as follows:

[0089] 1. Obtain recent cross-camera pedestrian re-identification records for all monitoring device nodes within the monitoring platform, and construct a pedestrian route network based on these records. Taking monitoring device nodes A and B as an example, assuming pedestrian α is detected when passing through monitoring device node A, and then detected again by monitoring device node B and identified as pedestrian α, it can be confirmed that there is an A→B propagation relationship between nodes A and B.

[0090] 2. Based on the constructed pedestrian route network, establish a PageRank algorithm model for the propagation efficiency value of each monitoring device node. The established PageRank algorithm model is set as follows:

[0091]

[0092] Where N represents the number of monitoring device nodes in the platform, d is the damping coefficient (usually taken as an empirical value of 0.85), PR(j) represents the propagation efficiency value of monitoring device node j, S(j) represents the set of monitoring device nodes for pedestrian re-identification in the pedestrian route network, PR(v) represents the propagation efficiency value of monitoring device node v in S(j) before this iteration, and L(v) is the number of outgoing links of monitoring device node v, representing the number of monitoring device nodes that can be re-identified after a pedestrian leaves monitoring device node v;

[0093] 3. The propagation efficiency value of each monitoring device node is determined using the PageRank algorithm model. This model is then iteratively applied to stabilize the propagation efficiency value of each monitoring device node in the platform, ultimately yielding the propagation efficiency value for each node. Based on these propagation efficiency values, key monitoring device nodes are identified from the monitoring platform. Finally, based on the propagation efficiency values ​​of these key monitoring device nodes, real-time assessment and dynamic push notifications of important areas are implemented on the monitoring platform.

[0094] Taking a monitoring platform with six monitoring device nodes (A, B, C, D, E, and F) as an example, recent video footage was selected. Using a pedestrian re-identification system, historical pedestrian paths included: from node A to node B (hereinafter referred to as A→B), A→C, A→D, B→D, B→E, C→E, D→E, E→A, and F→B. Using the PageRank algorithm, the final PageRank values ​​for the six monitoring nodes were calculated iteratively: A (0.282), B (0.126), C (0.104), D (0.158), E (0.303), and F (0.025). Sorted by PageRank value, the priority order for monitoring device nodes is E, A, D, B, C, and F. If the first three are selected, then the areas within the monitoring range of device nodes E, A, and D should be the focus of attention.

[0095] For monitoring equipment in hotspot areas, customized processing can also be performed, such as recording video in hotspot areas, extending the video retention period, and notifying platform administrators, which can effectively improve the efficiency of real-time hotspot monitoring on the monitoring platform.

[0096] II. The specific steps for recommending device-linked videos during emergencies are as follows:

[0097] 1) First, extract the recent cross-camera pedestrian re-identification records of all monitoring device nodes in the monitoring platform to construct a pedestrian route network.

[0098] Based on the constructed pedestrian route network, a PersonalRank algorithm model is established for the relevance value of a specific monitoring device node. The established PersonalRank algorithm model is defined as follows:

[0099]

[0100] Where u represents the selected monitoring device node, and it is necessary to recommend monitoring nodes with strong correlation. The above formula is used to calculate the correlation between all monitoring device nodes and the current monitoring device node u.

[0101] 2) To calculate the relevance of all monitoring device nodes relative to monitoring device node u, PersonalRank starts by iterating through the nodes corresponding to monitoring device node u until it stabilizes, ultimately obtaining the propagation efficiency value of the current monitoring device node u relative to all other monitoring device nodes. This propagation efficiency value can then be used for relevance ranking.

[0102] Taking a monitoring platform with six monitoring device nodes (A, B, C, D, E, and F) as an example, the recent pedestrian historical paths include A→B, A→C, A→D, B→D, B→E, C→E, D→E, E→A, and F→B. Now, we need to observe monitoring device node A and its neighboring nodes. Using the PersonalRank algorithm, the final PR values ​​for the other five monitoring nodes are B (0.1610), C (0.1029), D (0.1345), E (0.0611), and F (0.0684). Based on the PR value ranking, the relevance to monitoring device node A is in the order of B, D, C, F, and E. Therefore, when focusing on the monitoring area of ​​device node A, we should also prioritize observing the situations of device nodes B, D, and C.

[0103] Calculations show that device nodes B, D, and C are the closest and most relevant to device node A. Assuming a sudden incident occurs in the monitoring area of ​​device node A requiring close monitoring, the subsequent impact is most likely to affect areas B, D, and C. After selecting device node A for monitoring, the monitoring platform can automatically recommend video feeds from areas B, D, and C to monitoring personnel. This reduces management time, increases the ability to focus on key monitoring areas, and improves the automation level of the monitoring platform.

[0104] Based on the above solutions, the following are some practical scenarios for illustration:

[0105] A selected monitoring platform, after classification, comprises 105 monitoring nodes across 6 regions, with interconnected paths between regions. For ease of description, the monitoring platform is simplified to A through F, and the monitoring device nodes are designated A1 through F8, as shown in the table below:

[0106] Table 1. Information on nodes in each region of the monitoring platform

[0107] area Number of nodes A 29 B 16 C 11 D 22 E 19 F 8

[0108] Cross-camera pedestrian records over a period of time are extracted, and device videos in hotspot areas are recommended in real time. The cross-camera pedestrian data includes pedestrian re-identification results within and across regions, totaling 44 records. The cross-camera records are substituted into the PageRank model in the form of ("A7","A10"), ("A24","B15"), ("B13","E10")... After 29 rounds of iteration, the values ​​stabilize, yielding PageRank values ​​of A5 (0.0541), A10 (0.0495), B7 (0.0445), B9 (0.0402), B16 (0.0398), B15 (0.0391),..., F7 (0.004). The cross-camera record information is then compared based on the PageRank values.

[0109] Table 2 Monitoring Node Inbound and Outbound Information Table

[0110] Serial Number node Inbound Inbound nodes Out-of-chain Outgoing nodes 1 A5 2 A10, A15 3 A7, A10, B15 2 A10 3 A5, A7, A24 1 A5 3 B7 2 B9, B15 2 B9, B15 4 B9 2 B7, B12 2 B7, B12 5 B16 2 B12, E10 2 B12, E10 6 B15 2 A5, B7 2 A24, B7 … … … … … … 105 F7 0 — 0 —

[0111] As shown in the table above, nodes with relatively high numbers of incoming and outgoing chains have larger PR values ​​and are generally considered to be hotspot nodes, while nodes with zero incoming and outgoing chains have the smallest PR values ​​and can better reflect real-time hotspot areas within the monitoring platform.

[0112] For selecting video footage associated with devices during emergencies, taking the current period as an example, we input cross-camera recordings into the PersonalRank model and select node B15 as the node of interest. The resulting PR values ​​are B7, A24, A15, C7, B9, A5, ..., F8. The association between these nodes and B15 is shown in the table below:

[0113] Table 3. Information on associated nodes of the monitoring platform

[0114] Serial Number node PR value Shortest path 1 B7 0.1709 B15→B7 2 A24 0.1486 B15→A24 3 A15 0.0880 A15→A5→B15 4 C7 0.0758 C7→B15 5 B9 0.0647 B15→B7→B9 6 A5 0.0586 A5→B15 … — — — 104 F8 0.0 none

[0115] As shown in the table above, in the PersonalRank model, nodes with associated paths to node B15 have higher PR values, and the nodes directly pointed to by B15's outgoing chains have the highest PR values. This model can effectively recommend monitoring devices associated with nodes that have been followed to the monitoring platform, which has strong practical significance.

[0116] It should be noted that existing monitoring platforms manage multi-channel real-time video playback through pre-selected monitoring devices. Compared with existing technologies, the cross-camera video recommendation method provided in this application uses a cross-camera pedestrian re-identification algorithm to evaluate important nodes within the monitoring area. It can also evaluate related devices based on specified monitoring device nodes, reducing the workload of monitoring personnel and greatly improving management efficiency.

[0117] The cross-camera video recommendation device provided in the embodiments of this application is described below. The cross-camera video recommendation device described below can be referred to in correspondence with the cross-camera video recommendation method described above.

[0118] Please see Figure 2 This application provides a cross-camera video recommendation device, including:

[0119] Network construction module 1 is used to construct a pedestrian route network based on the cross-camera pedestrian re-identification records of all monitoring device nodes in the monitoring platform;

[0120] The propagation efficiency calculation module 2 is used to calculate the propagation efficiency value of each monitoring device node based on the pedestrian route network and using a pre-built node propagation efficiency algorithm model.

[0121] The critical node determination module 3 is used to determine the monitoring device nodes whose propagation efficiency values ​​meet the preset first condition as the critical monitoring device nodes of the monitoring platform.

[0122] Hotspot area monitoring module 4 is used to execute hotspot area real-time monitoring strategies on the key monitoring device nodes.

[0123] In one embodiment, the cross-camera video recommendation device further includes:

[0124] The correlation calculation module is used to calculate the correlation between each monitoring device node and the target monitoring device node based on the pedestrian route network and using a pre-built node correlation algorithm model when the target monitoring device node triggers a sudden event monitoring request.

[0125] The associated node determination module is used to determine the monitoring device nodes whose relevance meets the preset second condition as the associated monitoring device nodes of the target monitoring device node;

[0126] The associated area monitoring module is used to execute emergency monitoring strategies for sudden events on the associated monitoring device nodes.

[0127] In one embodiment, the hotspot area monitoring module 4 is specifically used for:

[0128] The monitoring cycle and video recording duration of the corresponding key monitoring equipment nodes are determined based on the magnitude of each propagation efficiency value.

[0129] Based on a defined monitoring cycle, dynamic video feeds are pushed to the areas corresponding to each key monitoring device node, and video recordings and storage are performed on the areas corresponding to each key monitoring device node based on a defined video recording duration.

[0130] In one embodiment, the associated area monitoring module is specifically used for:

[0131] The monitoring cycle and video recording duration of the corresponding associated monitoring device nodes are determined based on the magnitude of each relevant value.

[0132] Based on a defined monitoring cycle, dynamic video feeds are pushed to the areas corresponding to each associated monitoring device node, and video recordings and storage are performed on the areas corresponding to each associated monitoring device node based on a defined video recording duration.

[0133] In one embodiment, the key node determination module 3 is specifically used for:

[0134] All monitoring device nodes are sorted in descending order of propagation efficiency values, and the top-ranked nodes are identified as key monitoring device nodes for the monitoring platform; or,

[0135] Monitoring device nodes whose propagation efficiency values ​​are greater than a preset first threshold are identified as key monitoring device nodes of the monitoring platform.

[0136] In one embodiment, the associated node determination module is specifically used for:

[0137] All monitoring device nodes are sorted in descending order of relevance, and the top-ranked nodes are identified as associated monitoring device nodes of the target monitoring device node; or,

[0138] Monitoring device nodes with a relevance greater than a preset second threshold are identified as associated monitoring device nodes of the target monitoring device node.

[0139] In one embodiment, the node propagation efficiency algorithm model is the PageRank algorithm model, and the node correlation algorithm model is the PersonalRank algorithm model.

[0140] It is understood that the above-described device embodiments correspond to the method embodiments of this application. The cross-camera video recommendation device provided in this application can implement the cross-camera video recommendation method provided in any one of the method embodiments of this application.

[0141] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include a processor 310, a communication interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communication interface 320, and the memory 330 communicate with each other via the communication bus 340. The processor 310 can call a computer program in the memory 330 to execute steps of a cross-camera video recommendation method, such as including:

[0142] S1. Construct a pedestrian route network based on the cross-camera pedestrian re-identification records of all monitoring device nodes in the monitoring platform;

[0143] S2. Based on the pedestrian route network, calculate the propagation efficiency value of each monitoring device node using a pre-built node propagation efficiency algorithm model;

[0144] S3. The monitoring device nodes whose propagation efficiency values ​​meet the preset first condition are identified as the key monitoring device nodes of the monitoring platform.

[0145] S4. Implement a real-time monitoring strategy for hotspot areas on the key monitoring equipment nodes.

[0146] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0147] On the other hand, this application also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can perform the steps of the cross-camera video recommendation method provided in the above embodiments, such as including:

[0148] S1. Construct a pedestrian route network based on the cross-camera pedestrian re-identification records of all monitoring device nodes in the monitoring platform;

[0149] S2. Based on the pedestrian route network, calculate the propagation efficiency value of each monitoring device node using a pre-built node propagation efficiency algorithm model;

[0150] S3. The monitoring device nodes whose propagation efficiency values ​​meet the preset first condition are identified as the key monitoring device nodes of the monitoring platform.

[0151] S4. Implement a real-time monitoring strategy for hotspot areas on the key monitoring equipment nodes.

[0152] On the other hand, embodiments of this application also provide a processor-readable storage medium storing a computer program for causing a processor to perform the steps of the methods provided in the above embodiments, such as including:

[0153] S1. Construct a pedestrian route network based on the cross-camera pedestrian re-identification records of all monitoring device nodes in the monitoring platform;

[0154] S2. Based on the pedestrian route network, calculate the propagation efficiency value of each monitoring device node using a pre-built node propagation efficiency algorithm model;

[0155] S3. The monitoring device nodes whose propagation efficiency values ​​meet the preset first condition are identified as the key monitoring device nodes of the monitoring platform.

[0156] S4. Implement a real-time monitoring strategy for hotspot areas on the key monitoring equipment nodes.

[0157] The processor-readable storage medium can be any available medium or data storage device that the processor can access, including but not limited to magnetic memory (e.g., floppy disk, hard disk, magnetic tape, magneto-optical disk (MO)), optical memory (e.g., CD, DVD, BD, HVD), and semiconductor memory (e.g., ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)).

[0158] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0159] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0160] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A cross-camera video recommendation method, characterized in that, include: Based on the cross-camera pedestrian re-identification records of all monitoring device nodes in the monitoring platform, a pedestrian route network is constructed. The pedestrian route network includes multiple monitoring device nodes, the route relationships between the monitoring device nodes, and the corresponding pedestrian re-identification probabilities. Based on the pedestrian route network, the propagation efficiency value of each monitoring device node is calculated using a pre-constructed node propagation efficiency algorithm model; the node propagation efficiency algorithm model adopts the PageRank algorithm model. The monitoring device nodes whose propagation efficiency values ​​meet the preset first condition are identified as the key monitoring device nodes of the monitoring platform. A real-time monitoring strategy for hotspot areas is implemented on the key monitoring equipment nodes.

2. The cross-camera video recommendation method according to claim 1, characterized in that, Also includes: When a target monitoring device node triggers a sudden event monitoring request, the correlation between each monitoring device node and the target monitoring device node is calculated based on the pedestrian route network using a pre-built node correlation algorithm model. The node correlation algorithm model adopts the PersonalRank algorithm model; The monitoring device node whose relevance meets the preset second condition is determined as the associated monitoring device node of the target monitoring device node; Implement emergency monitoring strategies for sudden events on the associated monitoring device nodes.

3. The cross-camera video recommendation method according to claim 1, characterized in that, The implementation of the hotspot area real-time monitoring strategy for the key monitoring device nodes includes: The monitoring cycle and video recording duration of the corresponding key monitoring equipment nodes are determined based on the magnitude of each propagation efficiency value. Based on a defined monitoring cycle, dynamic video feeds are pushed to the areas corresponding to each key monitoring device node, and video recordings and storage are performed on the areas corresponding to each key monitoring device node based on a defined video recording duration.

4. The cross-camera video recommendation method according to claim 2, characterized in that, The execution of emergency monitoring strategies for the associated monitoring device nodes includes: The monitoring cycle and video recording duration of the corresponding associated monitoring device nodes are determined based on the magnitude of each relevant value. Based on a defined monitoring cycle, dynamic video feeds are pushed to the areas corresponding to each associated monitoring device node, and video recordings and storage are performed on the areas corresponding to each associated monitoring device node based on a defined video recording duration.

5. The cross-camera video recommendation method according to claim 1, characterized in that, The step of identifying monitoring device nodes whose propagation efficiency values ​​meet a preset first condition as key monitoring device nodes of the monitoring platform includes: All monitoring device nodes are sorted in descending order of propagation efficiency values, and the top-ranked nodes are identified as key monitoring device nodes for the monitoring platform; or, Monitoring device nodes whose propagation efficiency values ​​are greater than a preset first threshold are identified as key monitoring device nodes of the monitoring platform.

6. The cross-camera video recommendation method according to claim 2, characterized in that, The step of determining the monitoring device nodes whose relevance meets the preset second condition as the associated monitoring device nodes of the target monitoring device node includes: All monitoring device nodes are sorted in descending order of relevance, and the top-ranked nodes are identified as associated monitoring device nodes of the target monitoring device node; or, Monitoring device nodes with a relevance greater than a preset second threshold are identified as associated monitoring device nodes of the target monitoring device node.

7. A video recommendation device that supports multiple cameras, characterized in that, include: The network construction module is used to construct a pedestrian route network based on the cross-camera pedestrian re-identification records of all monitoring device nodes in the monitoring platform; The pedestrian route network includes multiple monitoring device nodes, the route relationships between the monitoring device nodes, and the corresponding pedestrian re-identification probabilities. The propagation efficiency calculation module is used to calculate the propagation efficiency value of each monitoring device node based on the pedestrian route network and using a pre-built node propagation efficiency algorithm model; the node propagation efficiency algorithm model adopts the PageRank algorithm model. The critical node determination module is used to determine the monitoring device nodes whose propagation efficiency values ​​meet a preset first condition as the critical monitoring device nodes of the monitoring platform. The hotspot area monitoring module is used to execute real-time hotspot area monitoring strategies on the key monitoring device nodes.

8. An electronic device comprising a processor and a memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the cross-camera video recommendation method according to any one of claims 1 to 6.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the cross-camera video recommendation method according to any one of claims 1 to 6.

Images