Monitoring resource scheduling method and system of intelligent video monitoring system

Abstract

The invention relates to a monitoring resource scheduling method and system of an intelligent video monitoring system, and relates to the field of resource scheduling, historical monitoring videos of monitoring equipment are called, abnormal personnel and abnormal events are respectively identified to obtain a plurality of abnormal personnel and event sequences, personnel continuous anomaly degree analysis is carried out to obtain personnel continuous anomaly degrees, and the personnel continuous anomaly degrees are analyzed to obtain a plurality of abnormal event sequences. Performing event continuous anomaly analysis according to the event chain library to obtain event continuous anomaly, and performing anomaly density analysis according to the plurality of abnormal persons and the event sequence to obtain a plurality of anomaly densities, allocating and scheduling storage resources of the plurality of monitoring devices in the video storage server according to the continuous anomalies of the plurality of persons, the continuous anomalies of the plurality of events and the density of the plurality of anomalies to obtain a plurality of storage resources as monitoring resource scheduling results; the technical problems of unreasonable video monitoring storage resource allocation, inaccurate abnormal event detection and low data storage efficiency are solved.

Description

Technical Field

[0001] The present invention relates to the field of resource scheduling, and particularly to a method and system for monitoring resource scheduling of an intelligent video monitoring system. Background Art

[0002] With the rapid development and popularization of monitoring technologies, intelligent video monitoring systems play an increasingly important role in fields such as social security, traffic management, and industrial production. However, existing video monitoring systems often face bottleneck problems such as limited storage resources, inaccurate detection of abnormal events, and low data storage efficiency.

[0003] Existing video monitoring systems usually adopt a fixed storage strategy, that is, regardless of the actual situation in the monitoring scenario, storage is performed at fixed time intervals or based on video size. This method not only wastes storage resources but may also cause video data of critical abnormal events to be overwritten or missed, thus unable to provide effective support for subsequent abnormal analysis and event tracing. Secondly, most abnormal event detection methods are based on simple image processing and rule matching algorithms, and their accuracy and robustness are often affected by factors such as the complexity of the monitoring scenario, light changes, and occlusion. Therefore, in practical applications, existing abnormal event detection methods are prone to false alarms and missed alarms, and cannot meet the requirements for accurate analysis and processing of monitoring data. Moreover, with the continuous increase in the number of monitoring devices and the continuous improvement of the resolution of monitoring videos, the amount of data generated by the monitoring system is showing an explosive growth. Existing storage and transmission methods can no longer meet the requirements for efficient storage and fast transmission of large-scale video monitoring data, and there is an urgent need for a more intelligent and efficient monitoring resource scheduling method to address this challenge. Summary of the Invention

[0004] In view of the technical problems in the prior art of unreasonable allocation of video monitoring storage resources, inaccurate detection of abnormal events, and low data storage efficiency, the present invention provides a method and system for monitoring resource scheduling of an intelligent video monitoring system to solve these problems.

[0005] The technical solutions of the present invention for solving the above technical problems are as follows:

[0006] In a first aspect, the present invention provides a monitoring resource scheduling method for an intelligent video monitoring system. The method includes: retrieving multiple historical monitoring videos of multiple monitoring devices through a video storage server in the video monitoring system, wherein the multiple monitoring devices are connected to the video storage server and send the monitoring videos to the video storage server for updated storage; respectively performing abnormal person and abnormal event recognition based on the multiple historical monitoring videos to obtain multiple abnormal person sequences and multiple abnormal event sequences; performing continuous abnormal degree analysis of persons based on the multiple abnormal person sequences to obtain multiple continuous abnormal degrees of persons, performing continuous abnormal degree analysis of events according to an event chain library based on the multiple abnormal event sequences to obtain multiple continuous abnormal degrees of events, and performing abnormal density analysis based on the multiple abnormal person sequences and multiple abnormal event sequences to obtain multiple abnormal densities; and performing allocation and scheduling of the storage resources of the multiple monitoring devices in the video storage server based on the multiple continuous abnormal degrees of persons, multiple continuous abnormal degrees of events, and multiple abnormal densities to obtain multiple storage resources as the monitoring resource scheduling result.

[0007] In a second aspect, the present invention provides a monitoring resource scheduling system for an intelligent video monitoring system. The system includes: a video retrieval module for retrieving multiple historical monitoring videos of multiple monitoring devices through a video storage server in the video monitoring system, wherein the multiple monitoring devices are connected to the video storage server and send the monitoring videos to the video storage server for updated storage; an abnormal recognition module for respectively performing abnormal person and abnormal event recognition based on the multiple historical monitoring videos to obtain multiple abnormal person sequences and multiple abnormal event sequences; an abnormal degree analysis module for performing continuous abnormal degree analysis of persons based on the multiple abnormal person sequences to obtain multiple continuous abnormal degrees of persons, performing continuous abnormal degree analysis of events according to an event chain library based on the multiple abnormal event sequences to obtain multiple continuous abnormal degrees of events, and performing abnormal density analysis based on the multiple abnormal person sequences and multiple abnormal event sequences to obtain multiple abnormal densities; and an allocation and scheduling module for performing allocation and scheduling of the storage resources of the multiple monitoring devices in the video storage server based on the multiple continuous abnormal degrees of persons, multiple continuous abnormal degrees of events, and multiple abnormal densities to obtain multiple storage resources as the monitoring resource scheduling result.

[0008] The beneficial effects of the present invention are: by intelligently analyzing abnormal persons and events in historical monitoring videos and combining the continuous abnormal degree of persons, continuous abnormal degree of events, and abnormal density, precise scheduling and optimized allocation of the storage resources of monitoring devices are achieved, effectively improving the utilization rate of monitoring resources and the accuracy of abnormal event detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1Flow schematic diagram of a monitoring resource scheduling method for an intelligent video monitoring system provided by the present invention.

[0010] Figure 2 Structural schematic diagram of a monitoring resource scheduling system for an intelligent video monitoring system provided by the present invention.

[0011] Explanation of reference numerals: video retrieval module 11, anomaly recognition module 12, anomaly degree analysis module 13, allocation and scheduling module 14. Detailed implementation manners

[0012] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present invention.

[0013] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, "a plurality" means two or more, unless otherwise specifically defined.

[0014] In the description of the present invention, the term "for example" is used to mean "serving as an example, illustration, or explanation". Any embodiment described as "for example" in the present invention is not necessarily construed as being more preferred or having more advantages than other embodiments. In order for any person skilled in the art to implement and use the present invention, the following description is given. In the following description, details are set forth for purposes of explanation. It should be understood that those of ordinary skill in the art can recognize that the present invention can be implemented without these specific details. In other instances, well-known structures and processes are not elaborated in detail to avoid unnecessary details from obscuring the description of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope that conforms to the principles and features disclosed in the present invention.

[0015] Embodiment 1:

[0016] As Figure 1 shown, the embodiment of the present invention provides a monitoring resource scheduling method for an intelligent video monitoring system, and the method includes:

[0017] S10: Retrieve multiple historical surveillance videos of multiple surveillance devices through the video storage server within the video surveillance system. Among them, the multiple surveillance devices are connected to the video storage server, and the surveillance videos are sent to the video storage server for update and storage.

[0018] Exemplarily, a video surveillance system refers to an entire deployment device that can collect image information of a surveillance area in real time through devices such as front-end cameras, and transmit this information to a surveillance center through a transmission system. At the surveillance center, staff can view the images of the surveillance area in real time through display devices and handle abnormal situations in a timely manner. The video surveillance system mainly includes parts such as front-end cameras, transmission systems, surveillance centers, and abnormal situation identification and early warning. Among them, front-end cameras are installed in key areas that need to be monitored, such as streets, shopping malls, etc. They can capture images of the surveillance area in real time and convert them into digital signals for transmission. The image information captured by the front-end cameras needs to be transmitted to the surveillance center through the transmission system. At the surveillance center, staff can view the images captured by the front-end cameras in real time through display devices (such as large-screen displays, video walls, etc.). At the same time, the surveillance center is also equipped with facilities such as video recorders and storage devices for saving and managing surveillance videos. At the same time, the video surveillance system also has an abnormal situation identification function. For example, through artificial intelligence technology, abnormal behaviors in the surveillance images (such as people breaking in, items missing, etc.) can be automatically identified and the alarm mechanism can be automatically triggered to notify staff to handle it in a timely manner. In short, the video surveillance system is applicable in multiple scenarios, such as public security, traffic management, industrial production, etc. It provides an effective means of security prevention for people by capturing and transmitting surveillance images in real time.

[0019] In this solution, specifically, the historical surveillance videos of multiple surveillance devices are retrieved through the video storage server. These surveillance devices are connected to the video storage server and can send the captured surveillance videos to the server for storage and update in real time or at regular intervals. For example, the system may be set to retrieve the surveillance videos within the last 3 days as the analysis object. These video data cover the images of different time periods and surveillance areas and contain rich information. Through such a process, a large number of historical surveillance videos can be obtained as the basis for subsequent abnormal situation identification and storage resource scheduling.

[0020] S20: Identify abnormal personnel and abnormal events respectively based on the multiple historical surveillance videos to obtain multiple abnormal personnel sequences and multiple abnormal event sequences.

[0021] Optionally, abnormal persons and abnormal events are identified based on multiple retrieved historical surveillance videos. Specifically, video recognition technology is used to analyze and process the image information in the surveillance footage. For the identification of abnormal persons, characteristic information of the persons (such as facial features, body shape features, etc.) can be captured, and through specific algorithms, comparison and marking are carried out, and a unique number or sequence is assigned to each identified abnormal person. At the same time, the system can also identify abnormal events in the surveillance footage, such as loitering events (persons staying or loitering in the surveillance area for a long time), vandalism events (persons or objects causing damage to the surveillance area), etc. These abnormal events will also be recorded and corresponding abnormal event sequences will be generated. For example, if there is an abnormal behavior of a person loitering in a certain surveillance footage, the system can automatically identify and record it, and at the same time mark the person as an abnormal person, and generate corresponding abnormal person sequences and abnormal event sequences to provide data support for subsequent analysis and processing.

[0022] S30: Perform continuous abnormality analysis of persons based on the multiple abnormal person sequences to obtain multiple continuous abnormality degrees of persons. Perform continuous abnormality analysis of events according to the event chain library based on the multiple abnormal event sequences to obtain multiple continuous abnormality degrees of events. Based on the multiple abnormal person sequences and multiple abnormal event sequences, perform abnormal density analysis to obtain multiple abnormal densities.

[0023] Furthermore, in-depth analysis of multiple abnormal person sequences is carried out to calculate the continuous abnormality degree of each person. The core of this process lies in identifying abnormal persons who frequently or continuously appear in the surveillance footage and assessing the risk that they may carry out abnormal behaviors. If the same abnormal person appears repeatedly in different time periods or different surveillance areas, the system will consider that this person has a high continuous abnormality degree, which may mean that they are preparing to carry out some illegal behavior, such as theft, etc. Through the analysis of the appearance frequency, behavior patterns of these abnormal persons and their correlation with other surveillance data, the continuous abnormality degree of each person can be quantitatively evaluated and corresponding data can be generated. For example, if a certain abnormal person appears in the surveillance footage of a certain commercial area multiple times in the past few days and is accompanied by suspicious behaviors each time, then the system will mark him as a highly continuously abnormal person and assign a high continuous abnormality degree to him. This analysis is crucial for subsequent storage resource scheduling, because the greater the continuous abnormality degree of a person, the greater the demand for retrieving past surveillance videos when illegal events occur. Therefore, more storage resources should be allocated to ensure that key evidence is not missed.

[0024] Subsequently, further analysis is performed on multiple abnormal event sequences to calculate the event consecutive abnormality degree. This step mainly conducts correlation analysis and pattern matching on the abnormal events occurring in the monitoring based on a preset event chain library. The event chain library contains various known combinations and occurrence orders of abnormal events, and these combinations and orders are often closely related to illegal acts. For example, a typical abnormal event chain might be "observing - lingering - approaching public facilities", and such an event combination might imply that someone is attempting to carry out illegal acts such as theft in accordance with this chain.

[0025] Next, each abnormal event sequence is disassembled and compared to find the event combinations that match the event chain library, and the proportion of these combinations in the total event sequence is calculated. This proportion is defined as the event consecutive abnormality degree. For example, if three abnormal events of observing, lingering, and approaching public facilities occur continuously in a certain monitoring area within a short period of time, and the order of these events is consistent with a certain chain in the event chain library, then the system will consider this event sequence to have a high consecutive abnormality degree. The calculation of the event consecutive abnormality degree is also of great significance for subsequent storage resource scheduling. The greater the event consecutive abnormality degree, the higher the likelihood of illegal events occurring, and the greater the need to schedule past monitoring videos. In this case, the system needs to allocate more storage resources to save the information within these key event chains for subsequent analysis and investigation, which can not only improve the response speed and accuracy of the monitoring system but also provide strong evidence support for relevant departments.

[0026] After that, abnormal density analysis is carried out based on multiple abnormal person sequences and multiple abnormal event sequences. This analysis process mainly counts the number of abnormal persons and abnormal events within a specific time period or monitoring area, and calculates the abnormal density accordingly. Abnormal density is a quantitative indicator used to reflect the frequency of abnormal activities in the monitoring area. Specifically, if there are more abnormal persons and more abnormal events in a certain monitoring area, then the abnormal density of this area is greater. For example, in the monitoring footage of a commercial area, if multiple abnormal persons linger and gather within a short period of time, and abnormal events such as item loss and facility damage occur, then the system will calculate that the abnormal density of this area is high. The analysis result of abnormal density has important guiding significance for the allocation of storage resources. The greater the abnormal density, the more important the monitoring data of this area, and the greater the need to schedule and view past monitoring videos. Therefore, the system will allocate different storage resources for different monitoring areas or time periods according to the size of the abnormal density to ensure that key monitoring data can be properly saved and efficiently utilized.

[0027] S40: Based on the continuous abnormality degrees of multiple personnel, the continuous abnormality degrees of multiple events, and multiple abnormality densities, perform allocation and scheduling of the storage resources of the multiple monitoring devices in the video storage server to obtain multiple storage resources as the monitoring resource scheduling result.

[0028] Specifically, perform comprehensive storage resource allocation and scheduling according to the calculated continuous abnormality degrees of multiple personnel, the continuous abnormality degrees of multiple events, and multiple abnormality densities. The core of this process lies in integrating these three key indicators to comprehensively evaluate the importance of the monitoring area and the frequency of abnormal activities. Specifically, take the continuous abnormality degree of personnel, the continuous abnormality degree of events, and the abnormality density as input parameters, and perform comprehensive analysis of weight coefficient allocation through intelligent algorithms to determine the amount of storage resources that should be allocated to each monitoring device in the video storage server. For example, if the continuous abnormality degree of personnel and the continuous abnormality degree of events in a certain monitoring area are both high, and the abnormality density is also large, then the system will consider the monitoring data in this area to be very important and needs to allocate more storage resources to ensure the integrity and traceability of the data. On the contrary, if these three indicators in a certain monitoring area are all low, then the system will reduce the storage resource allocation for this area. Through such allocation and scheduling, the utilization efficiency of storage resources can be optimized, ensuring that key monitoring data is properly stored while reducing storage costs. Finally, the system will allocate corresponding storage resources to each monitoring device according to this comprehensive evaluation result and implement it as the final result of monitoring resource scheduling.

[0029] In a preferred embodiment, through the video storage server in the video monitoring system, retrieve multiple historical monitoring videos of multiple monitoring devices, including: start a resource scheduling instruction according to a preset resource scheduling period; in response to the resource scheduling instruction, retrieve the monitoring videos of multiple monitoring devices within a preset time range in the past in the video storage server to obtain multiple historical monitoring videos.

[0030] Preferably, in order to effectively manage and utilize storage resources, the system will automatically start a resource scheduling instruction according to a preset resource scheduling period. The triggering of this instruction is periodic, aiming to ensure the timeliness and effectiveness of monitoring data. Once the resource scheduling instruction is started, the video storage server will immediately respond and retrieve the monitoring videos of multiple monitoring devices within a preset time range in the past according to the instruction requirements. The preset time range here can be set according to actual needs, such as the past week, month, or longer. Through such operations, historical monitoring videos of multiple monitoring devices can be obtained, and these video data will serve as the basis for subsequent abnormality identification, storage resource allocation and scheduling, etc. For example, if the system is set to perform resource scheduling at 0:00 on Monday every week, then every Monday at 0:00, the video storage server will automatically retrieve the monitoring videos of each monitoring device in the past week for subsequent analysis and processing.

[0031] In a preferred embodiment, based on the multiple historical surveillance videos, abnormal personnel and abnormal events are respectively identified to obtain multiple abnormal personnel sequences and multiple abnormal event sequences, including: respectively performing image segmentation on the multiple historical surveillance videos, combining every K adjacent surveillance images as a surveillance image group to obtain a set of multiple historical surveillance image groups, where K is a positive integer; respectively inputting each historical surveillance image group in the set of multiple historical surveillance image groups into the abnormal personnel recognition branch and the abnormal event recognition branch in a pre-trained abnormal recognizer, and recognizing and outputting the abnormal personnel and abnormal events in each historical surveillance image group; arranging the recognized abnormal personnel and abnormal events in the order of the time of the historical surveillance image groups in each set of historical surveillance image groups to obtain multiple abnormal personnel sequences and multiple abnormal event sequences.

[0032] In a specific embodiment, the identification of abnormal personnel and abnormal events is performed on multiple historical surveillance videos. First, image segmentation is performed on these historical surveillance videos. Specifically, every K adjacent surveillance images are combined into a surveillance image group, thereby obtaining a set of multiple historical surveillance image groups. Here, K is a positive integer, which represents the number of images included in each surveillance image group. For example, if K is set to 90 and the frame rate of the surveillance video is 30 frames per second, then the surveillance images every 3 minutes will be combined into a surveillance image group.

[0033] Each historical surveillance image group is respectively input into a pre-trained abnormal recognizer. This abnormal recognizer includes two branches: an abnormal personnel recognition branch and an abnormal event recognition branch. Through these two branches, the system can identify the abnormal personnel and abnormal events in each surveillance image group.

[0034] Thereafter, the recognized abnormal personnel and abnormal events are arranged in the order of the time of the historical surveillance image groups in each set of historical surveillance image groups, thereby obtaining multiple abnormal personnel sequences and multiple abnormal event sequences. These sequences will serve as the basis for subsequent analysis of the continuous abnormality degree of personnel, the continuous abnormality degree of events, and the abnormality density analysis. For example, if a set of surveillance image groups contains surveillance image groups every 3 minutes within the past hour, then the abnormal personnel and abnormal events in each time period within this hour are identified based on these image groups and arranged in chronological order.

[0035] In a preferred embodiment, the pre-training step of the anomaly recognizer includes: using a convolutional neural network to construct an abnormal person recognition branch and an abnormal event recognition branch respectively; according to the monitoring data records, collecting a set of sample monitoring image groups, and annotating the abnormal persons and abnormal events in each sample monitoring image group to obtain a set of sample abnormal persons and a set of sample abnormal events; using the set of sample monitoring image groups as input features, and using the set of sample abnormal persons and the set of sample abnormal events as output features respectively, and performing supervised training and testing on the abnormal person recognition branch and the abnormal event recognition branch until the accuracy rate meets the convergence threshold; combining the converged abnormal person recognition branch and abnormal event recognition branch to obtain an anomaly recognizer.

[0036] Specifically, a deep learning technique, namely a convolutional neural network, is used to construct an abnormal person recognition branch and an abnormal event recognition branch respectively. These two branches are the core components of the anomaly recognizer and are responsible for identifying abnormal persons and abnormal events in the monitoring images respectively.

[0037] Among them, the monitoring data record refers to a series of data for recording, storing, and managing the video stream or image sequence captured by the monitoring camera in the video monitoring system. These data record the actual situation of the monitored area during a specific period and are the basis for subsequent anomaly recognition, event backtracking, evidence extraction, etc. The monitoring data record includes content such as timestamps, image / video content, camera information, and storage format. For example, in the monitoring system of a large shopping mall, the monitoring data record may include the monitoring videos of various areas such as entrances, exits, corridors, and stores. When an abnormal event (such as theft, fighting, etc.) occurs, security personnel can retrieve the monitoring data record of the relevant time period, quickly locate the location and time of the event, and view the specific monitoring images to obtain evidence.

[0038] According to the monitoring data records, a large set of sample monitoring image groups is collected. These sample image group sets cover various possible monitoring scenarios and abnormal situations. Then, the abnormal persons and abnormal events in each sample monitoring image group are manually annotated to obtain a set of sample abnormal persons and a set of sample abnormal events. These annotated data will be used as the "true values" or "labels" for training the anomaly recognizer.

[0039] During the training process, the set of sample monitoring image groups is used as input features, and the set of sample abnormal persons and the set of sample abnormal events are used as output features respectively to perform supervised training on the abnormal person recognition branch and the abnormal event recognition branch. By continuously adjusting the network parameters and optimizing the algorithm, these two branches can accurately identify the abnormal persons and abnormal events in the monitoring images. The training process will continue until the accuracy rates of both branches meet the preset convergence threshold.

[0040] Finally, the converged abnormal person recognition branch and abnormal event recognition branch are combined together to form a complete abnormal recognizer. This abnormal recognizer has the ability to identify abnormal persons and abnormal events in actual monitoring scenarios. For example, in a surveillance system of a shopping mall, the pre-trained abnormal recognizer can analyze surveillance videos in real time and accurately identify suspicious persons wandering or abnormal events occurring, such as item loss, people falling, etc., thus providing strong support for the security management of the shopping mall.

[0041] In a preferred embodiment, continuous abnormal degree analysis of personnel is performed based on the multiple abnormal person sequences to obtain multiple continuous abnormal degrees of personnel, including: screening out abnormal persons with repetitions within the multiple abnormal person sequences to obtain multiple numbers of repeated persons; obtaining the number of abnormal persons within the multiple abnormal person sequences to obtain multiple numbers of abnormal persons; and respectively calculating the ratios of the multiple numbers of repeated persons and the multiple numbers of abnormal persons to obtain multiple continuous abnormal degrees of personnel.

[0042] Furthermore, in order to evaluate the potential threat level of abnormal persons, continuous abnormal degree analysis of personnel is carried out. Specifically, abnormal persons with repetitions are screened out within multiple abnormal person sequences, and the number of these repeated persons is counted to obtain multiple numbers of repeated persons. At the same time, the total number of abnormal persons in these abnormal person sequences is obtained, that is, multiple numbers of abnormal persons. Next, the ratio of the number of repeated persons to the total number of abnormal persons in each abnormal person sequence is calculated respectively, and this ratio represents the continuous abnormal degree of this person in this sequence. Through this process, the continuous abnormal degree of each abnormal person can be quantitatively evaluated. For example, if a certain abnormal person appears frequently in multiple monitoring time periods, then the number of repeated persons will be relatively high, and the ratio to the total number of abnormal persons will also increase accordingly, indicating that this person has a relatively high continuous abnormal degree and may be implementing or preparing to implement some abnormal behavior, and the system can allocate more attention and storage resources to it.

[0043] For example, assume there is a video surveillance system in a shopping mall. The system captures multiple sequences of abnormal personnel within a day, and each sequence represents the situation of abnormal personnel in the mall during a specific time period. The data of abnormal personnel sequences is as follows: Sequence 1: A, B, C, A (Abnormal personnel A appears 2 times, and the total number of abnormal personnel is 4); Sequence 2: B, D, E, B, F (Abnormal personnel B appears 2 times, and the total number of abnormal personnel is 5); Sequence 3: A, G, H, I (Abnormal personnel A appears 1 time, and the total number of abnormal personnel is 4); Sequence 4: B, J, K, B, L, B (Abnormal personnel B appears 3 times, and the total number of abnormal personnel is 6). When calculating Sequence 1, the number of repeated personnel: A appears 2 times, so the number of repeated personnel is 1 (only calculate the number of different types of repeated personnel, without accumulating the number of times), and the corresponding number of repetitions is 2 (when calculating the ratio, use the highest number of repetitions corresponding to the type to reflect the fact of "existence of repetition". Here, it is simplified to use the repeated appearance of A to illustrate the repetition situation, and the ratio calculation uses 1 type of repeated personnel), that is, the number of abnormal personnel is 4, and the continuous abnormality degree of personnel is 1 / 4 = 0.25 (indicating that there are repeated personnel, and the proportion of the types of repeated personnel in the total number of abnormal personnel). Among them, if strictly calculated according to the number of appearances for the "influence of repetition degree", it can be understood that if A is counted multiple times, the "repetition influence value" will be higher, but here for simplicity of explanation, it is reflected by the number of types. Similarly, when calculating Sequence 2, the number of repeated personnel: B appears 2 times, so the number of repeated personnel is 1; the number of abnormal personnel: 5; the continuous abnormality degree of personnel: 1 / 5 = 0.2. Sequences 3 and 4 are in the same form. Then, for result analysis, the continuous abnormality degree of personnel in Sequence 1 is 0.25, indicating that within this time period, there are repeated abnormal personnel, and the proportion of the types of repeated personnel is relatively high; the continuous abnormality degree of personnel in Sequence 2 is 0.2, indicating that the proportion of repeated personnel is relatively low; the continuous abnormality degree of personnel in Sequence 3 is 0, indicating that there are no repeated abnormal personnel within this time period; although B appears 3 times in Sequence 4, when calculating the ratio, it is still calculated according to the types of repeated personnel, and its continuous abnormality degree of personnel is approximately 0.167, which is the lowest among the four sequences. Through the above process, the continuous abnormality degree of abnormal personnel can be quantitatively evaluated. In practical applications, the system can allocate storage resources, adjust monitoring strategies or perform other relevant security management tasks according to these continuous abnormality degrees of personnel.

[0044] In a preferred embodiment, event consecutive abnormality analysis is performed according to the multiple abnormal event sequences in the event chain library to obtain multiple event consecutive abnormality degrees, including: within the multiple abnormal event sequences, respectively combine the N abnormal events after each abnormal event to obtain multiple abnormal event group sets, where N is a positive integer; input the multiple abnormal event group sets into the event chain library for retrieval to obtain multiple event chain numbers, where the event chain library includes multiple sample abnormal event chains, and when an abnormal event group is consistent with any one of the sample abnormal event chains, the event chain number count is incremented by 1; respectively calculate the ratios of the multiple event chain numbers to the number of abnormal event groups in the multiple abnormal event group sets to obtain multiple event consecutive abnormality degrees.

[0045] Specifically, to evaluate the continuity and potential threat of abnormal events, event consecutive abnormality analysis is carried out. First, the system will respectively combine the N abnormal events after each abnormal event within the multiple abnormal event sequences (for example, N is set to 3 for convenience in explaining in combination with subsequent event chains), thereby obtaining multiple abnormal event group sets. Each abnormal event group represents a sequence of abnormal events that occur continuously within a specific time window. Next, these abnormal event group sets are input into the event chain library for retrieval. The event chain library contains multiple sample abnormal event chains, and these chains represent known abnormal behavior patterns. For example, "observing - wandering - approaching public equipment" is a sample abnormal event chain. When it is found that an abnormal event group is exactly the same as any one of the sample abnormal event chains in the event chain library, it is considered that the abnormal event group constitutes an abnormal event chain, and the event chain number count is incremented by 1. Finally, respectively calculate the ratio of the event chain number to the number of abnormal event groups in each abnormal event sequence, and this ratio is the event consecutive abnormality degree of the abnormal event sequence. Through this process, the continuity and potential threat degree of abnormal events can be quantitatively evaluated.

[0046] For example, assume that in a certain monitoring area, the system captures the following abnormal event sequence: Event A - Event B - Event C - Event D. The system will first combine abnormal event groups such as "Event A - Event B - Event C" and "Event B - Event C - Event D". Then, these abnormal event groups are compared with the sample abnormal event chains in the event chain library. If it is found that "Event A - Event B - Event C" is exactly the same as the sample abnormal event chain "observing - wandering - approaching public equipment", then the event chain number count is incremented by 1. Finally, calculate the ratio of the event chain number to the number of all abnormal event groups to obtain the event consecutive abnormality degree of the abnormal event sequence. This abnormality degree can help the system determine whether the abnormal events in this sequence constitute a continuous and potentially threatening behavior pattern.

[0047] In a preferred embodiment, based on the multiple sequences of abnormal persons and the multiple sequences of abnormal events, abnormal density analysis is performed to obtain multiple abnormal densities, including: calculating the total number of abnormal persons and the total number of abnormal events within the multiple sequences of abnormal persons and the multiple sequences of abnormal events to obtain multiple abnormal person numbers and multiple abnormal event numbers; respectively calculating the ratio of each abnormal person number to the mean of the multiple abnormal person numbers to obtain multiple person abnormal densities; respectively calculating the ratio of each abnormal event number to the mean of the multiple abnormal event numbers to obtain multiple event abnormal densities; and calculating and obtaining multiple abnormal densities based on the multiple person abnormal densities and the multiple event abnormal densities.

[0048] Exemplarily, in order to comprehensively evaluate the abnormal activity level of the monitored area, abnormal density analysis is carried out. This analysis process first calculates the total number of abnormal persons and the total number of abnormal events within the multiple sequences of abnormal persons and the multiple sequences of abnormal events, so as to obtain multiple abnormal person numbers and multiple abnormal event numbers. These numbers reflect the frequency of abnormal activities in different time periods or different monitored areas.

[0049] Respectively calculate the ratio of each abnormal person number to the mean of the multiple abnormal person numbers, and this ratio is the person abnormal density. Similarly, respectively calculate the ratio of each abnormal event number to the mean of the multiple abnormal event numbers to obtain the event abnormal density. The person abnormal density and the event abnormal density respectively quantify the distribution density of abnormal persons in the monitored area and the multiple times of the occurrence frequency of abnormal events relative to the average level.

[0050] Multiple abnormal densities are calculated and obtained based on the multiple person abnormal densities and the multiple event abnormal densities. This abnormal density is a comprehensive index that simultaneously considers the number of abnormal persons and the occurrence frequency of abnormal events, so as to more comprehensively reflect the abnormal activity level of the monitored area.

[0051] For example, assume there are two monitored areas A and B. In a period of time, area A captures 10 abnormal persons and 5 abnormal events, while area B captures 20 abnormal persons and 15 abnormal events. If the average number of abnormal persons in all monitored areas is 15 and the average number of abnormal events is 10, then the person abnormal density of area A is 10 / 15 = 0.67, and the event abnormal density is 5 / 10 = 0.5; while the person abnormal density of area B is 20 / 15 ≈ 1.33, and the event abnormal density is 15 / 10 = 1.5. Finally, the system will comprehensively calculate the abnormal densities of area A and area B based on these person abnormal densities and event abnormal densities, so as to help security personnel more intuitively understand the abnormal activity levels of different monitored areas and take targeted monitoring measures.

[0052] In a preferred embodiment, according to the multiple continuous abnormality degrees of personnel, the multiple continuous abnormality degrees of events, and the multiple abnormality densities, the allocation and scheduling of the storage resources of the multiple monitoring devices in the video storage server are performed to obtain multiple storage resources as the monitoring resource scheduling result, including: respectively calculating and obtaining multiple first allocation coefficients, multiple second allocation coefficients, and multiple third allocation coefficients according to the multiple continuous abnormality degrees of personnel, the multiple continuous abnormality degrees of events, and the multiple abnormality densities; calculating and obtaining multiple allocation and scheduling coefficients according to the multiple first allocation coefficients, the multiple second allocation coefficients, and the multiple third allocation coefficients; performing allocation and scheduling on the storage resources in the video storage server according to the multiple allocation and scheduling coefficients to obtain multiple storage resources, and updating and recording the monitoring videos stored by the multiple monitoring devices as the monitoring resource scheduling result.

[0053] Specifically, in order to efficiently utilize the storage resources of the video storage server, the allocation and scheduling of the storage resources are performed according to the multiple continuous abnormality degrees of personnel, the multiple continuous abnormality degrees of events, and the multiple abnormality densities. First, multiple first allocation coefficients, multiple second allocation coefficients, and multiple third allocation coefficients are respectively calculated and obtained according to these metrics. Specifically, a monitoring device with a higher continuous abnormality degree of personnel may obtain a higher first allocation coefficient, meaning that the abnormal behaviors of personnel captured by these devices are more frequent and require more storage resources to record; similarly, a monitoring device with a higher continuous abnormality degree of events will obtain a higher second allocation coefficient, indicating that the abnormal events occurring in these areas are more continuous and complex and also require more storage resources; while a monitoring device with a higher abnormality density will obtain a higher third allocation coefficient, reflecting that the abnormal activities in these areas are more intensive and also require more storage resources to ensure the complete recording of the monitoring videos. Specifically, a simple weighted summation method can be used to calculate the allocation and scheduling coefficient of each device, and the specific weights can be set according to the actual scenario. Then, multiple allocation and scheduling coefficients are calculated and obtained according to these allocation coefficients. The allocation and scheduling coefficient is a comprehensive index that comprehensively considers the influence of the continuous abnormality degree of personnel, the continuous abnormality degree of events, and the abnormality density on the storage resource requirements, so as to more accurately reflect the actual storage resource requirements of each monitoring device. The system will perform allocation and scheduling on the storage resources in the video storage server according to these allocation and scheduling coefficients. Specifically, corresponding storage resources are allocated to each monitoring device to ensure that they can continuously and stably record the monitoring videos. At the same time, the monitoring video storage of the monitoring devices is updated and recorded as the monitoring resource scheduling result. In this way, when it is necessary to retrieve the monitoring videos or perform abnormal analysis later, the system can quickly and accurately locate the required video data.

[0054] For example, assume there are two monitoring devices A and B. The continuous anomaly degree of personnel, continuous anomaly degree of events, and anomaly density of device A are all relatively high. Therefore, it obtains relatively high first, second, and third allocation coefficients, and further calculates a relatively high allocation scheduling coefficient. In contrast, these indicators of device B are relatively low, so the obtained allocation coefficients are also low. When allocating storage resources, the system will allocate more storage resources to device A to ensure that it can completely record abnormal activities in the monitoring area; while allocating fewer storage resources to device B to save storage costs. Through such allocation and scheduling of storage resources, the system can utilize storage resources more efficiently and improve the overall performance of the monitoring system.

[0055] A monitoring resource scheduling method for an intelligent video monitoring system provided by an embodiment of the present invention has at least the following technical effects:

[0056] 1. By comprehensively considering the continuous anomaly degree of personnel, continuous anomaly degree of events, and anomaly density, accurate allocation of storage resources for monitoring devices is achieved. This allocation method not only considers the frequency of abnormal personnel and events but also the density of abnormal activities, thus being able to more accurately reflect the actual storage resource requirements of each monitoring device, helping to avoid waste of storage resources and improve the utilization efficiency of storage resources.

[0057] 2. The anomaly recognizer is constructed using a convolutional neural network and optimized through pre-training, enabling efficient and accurate recognition of abnormal personnel and events in monitoring videos. At the same time, by converting the monitoring video into a group of monitoring images through image segmentation and combination techniques for batch processing, the efficiency of anomaly recognition is further improved. In addition, through combining abnormal events and comparing them with the event chain library, the analysis of the continuous anomaly degree of events can discover potential continuous abnormal behavior patterns, providing stronger support for security monitoring.

[0058] 3. Adaptively adjust the allocation of storage resources according to the actual operation of the monitoring system. By regularly activating the resource scheduling instruction, the latest monitoring video data can be obtained in real time, and the allocation of storage resources can be dynamically adjusted according to the anomaly recognition and analysis results. This adaptive scheduling method enables the monitoring system to flexibly respond to monitoring requirements in different scenarios, improving the flexibility and response speed of the monitoring system.

[0059] Embodiment 2:

[0060] As Figure 2 shown, based on the same inventive concept as the monitoring resource scheduling method for an intelligent video monitoring system provided in Embodiment 1, an embodiment of the present invention further provides a monitoring resource scheduling system for an intelligent video monitoring system, and the system includes:

[0061] A video retrieval module 11 is used to retrieve multiple historical surveillance videos of multiple surveillance devices through a video storage server in a video surveillance system. Among them, the multiple surveillance devices are connected to the video storage server, and the surveillance videos are sent to the video storage server for updated storage.

[0062] An anomaly recognition module 12 is used to respectively recognize abnormal persons and abnormal events based on the multiple historical surveillance videos, and obtain multiple abnormal person sequences and multiple abnormal event sequences.

[0063] An anomaly degree analysis module 13 is used to perform continuous anomaly degree analysis of personnel based on the multiple abnormal person sequences to obtain multiple continuous anomaly degrees of personnel, perform continuous anomaly degree analysis of events according to an event chain library based on the multiple abnormal event sequences to obtain multiple continuous anomaly degrees of events, and perform anomaly density analysis based on the multiple abnormal person sequences and multiple abnormal event sequences to obtain multiple anomaly densities.

[0064] An allocation and scheduling module 14 is used to perform allocation and scheduling of storage resources of the multiple surveillance devices in the video storage server based on the multiple continuous anomaly degrees of personnel, multiple continuous anomaly degrees of events, and multiple anomaly densities, and obtain multiple storage resources as the surveillance resource scheduling result.

[0065] Furthermore, the video retrieval module 11 is also used to execute the following steps:

[0066] Start a resource scheduling instruction according to a preset resource scheduling period; in response to the resource scheduling instruction, retrieve surveillance videos of multiple surveillance devices within a preset time range in the video storage server to obtain multiple historical surveillance videos.

[0067] Furthermore, the anomaly recognition module 12 is also used to execute the following steps:

[0068] Respectively perform image segmentation on the multiple historical surveillance videos, combine every K adjacent surveillance images as a surveillance image group to obtain a set of multiple historical surveillance image groups, where K is a positive integer; respectively input each historical surveillance image group in the set of multiple historical surveillance image groups into the abnormal person recognition branch and abnormal event recognition branch in a pre-trained anomaly recognizer to recognize and output abnormal persons and abnormal events in each historical surveillance image group; arrange the recognized abnormal persons and abnormal events in the time order of the historical surveillance image groups in each set of historical surveillance image groups to obtain multiple abnormal person sequences and multiple abnormal event sequences.

[0069] Furthermore, the anomaly recognition module 12 is also used to execute the following steps:

[0070] Using a convolutional neural network, an abnormal person recognition branch and an abnormal event recognition branch are respectively constructed; according to the monitoring data records, a set of sample monitoring image groups is collected, and the abnormal persons and abnormal events in each sample monitoring image group are labeled to obtain a set of sample abnormal persons and a set of sample abnormal events; using the set of sample monitoring image groups as input features, and using the set of sample abnormal persons and the set of sample abnormal events as output features respectively, the abnormal person recognition branch and the abnormal event recognition branch are respectively supervised and trained and tested until the accuracy rate meets the convergence threshold; the converged abnormal person recognition branch and abnormal event recognition branch are combined to obtain an abnormal recognizer.

[0071] Furthermore, the abnormality analysis module 13 is further configured to perform the following steps:

[0072] Screen for repeated abnormal persons within the multiple abnormal person sequences to obtain a plurality of duplicate person numbers; obtain the number of abnormal persons within the multiple abnormal person sequences to obtain a plurality of abnormal person numbers; calculate the ratios of the plurality of duplicate person numbers and the plurality of abnormal person numbers respectively to obtain a plurality of person continuous abnormality degrees.

[0073] Furthermore, the abnormality analysis module 13 is further configured to perform the following steps:

[0074] Within the multiple abnormal event sequences, combine the N abnormal events after each abnormal event respectively to obtain a set of multiple abnormal event groups, where N is a positive integer; input the set of multiple abnormal event groups into the event chain library for retrieval to obtain a plurality of event chain numbers, where the event chain library includes a plurality of sample abnormal event chains, and when the abnormal event group is consistent with any one of the sample abnormal event chains, the event chain number is counted as 1; calculate the ratios of the plurality of event chain numbers and the number of abnormal event groups within the set of multiple abnormal event groups respectively to obtain a plurality of event continuous abnormality degrees.

[0075] Furthermore, the abnormality analysis module 13 is further configured to perform the following steps:

[0076] Calculate the total number of abnormal persons and the total number of abnormal events within the multiple abnormal person sequences and the multiple abnormal event sequences to obtain a plurality of abnormal person numbers and a plurality of abnormal event numbers; calculate the ratios of each abnormal person number to the mean of the plurality of abnormal person numbers respectively to obtain a plurality of person abnormal densities; calculate the ratios of each abnormal event number to the mean of the plurality of abnormal event numbers respectively to obtain a plurality of event abnormal densities; calculate and obtain a plurality of abnormal densities according to the plurality of person abnormal densities and the plurality of event abnormal densities.

[0077] Furthermore, the allocation and scheduling module 14 is further configured to perform the following steps:

[0078] Calculate and obtain a plurality of first allocation coefficients, a plurality of second allocation coefficients, and a plurality of third allocation coefficients respectively according to the plurality of continuous abnormality degrees of personnel, the plurality of continuous abnormality degrees of events, and the plurality of abnormality densities; calculate and obtain a plurality of allocation scheduling coefficients according to the plurality of first allocation coefficients, the plurality of second allocation coefficients, and the plurality of third allocation coefficients; perform allocation scheduling on the storage resources in the video storage server according to the plurality of allocation scheduling coefficients to obtain a plurality of storage resources, and update and record the monitoring video storage of the plurality of monitoring devices as the monitoring resource scheduling result.

[0079] Through the foregoing detailed description of a monitoring resource scheduling method for an intelligent video monitoring system in this specification, those skilled in the art can clearly know a monitoring resource scheduling system for an intelligent video monitoring system in this embodiment. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description in the method part.

[0080] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A monitoring resource scheduling method for an intelligent video monitoring system, characterized in that: The method comprises: Retrieving multiple historical surveillance videos of multiple surveillance devices through a video storage server in a video surveillance system, wherein the multiple surveillance devices are connected to the video storage server, and the surveillance videos are sent to the video storage server for updating and storage; According to the multiple historical surveillance videos, respectively identify abnormal persons and abnormal events to obtain multiple abnormal person sequences and multiple abnormal event sequences; Performing personnel continuous abnormality degree analysis according to the multiple abnormal personnel sequences to obtain multiple personnel continuous abnormality degrees, performing event continuous abnormality degree analysis according to the multiple abnormal event sequences and the event chain library to obtain multiple event continuous abnormality degrees, and performing abnormal density analysis according to the multiple abnormal personnel sequences and the multiple abnormal event sequences to obtain multiple abnormal densities; According to the multiple continuous abnormality degrees of personnel, the multiple continuous abnormality degrees of events and the multiple abnormality densities, the storage resources of the multiple monitoring devices in the video storage server are allocated and scheduled to obtain multiple storage resources as monitoring resource scheduling results.

2. The monitoring resource scheduling method of the intelligent video monitoring system according to claim 1, characterized in that: Through the video storage server in the video surveillance system, multiple historical surveillance videos of multiple surveillance devices are retrieved, including: According to the preset resource scheduling cycle, start the resource scheduling instruction; In response to the resource scheduling instruction, the video storage server retrieves monitoring videos of multiple monitoring devices within a preset time range in the past to obtain multiple historical monitoring videos.

3. The monitoring resource scheduling method of the intelligent video monitoring system according to claim 1, characterized in that: According to the multiple historical surveillance videos, abnormal persons and abnormal events are identified respectively to obtain multiple abnormal person sequences and multiple abnormal event sequences, including: Performing image segmentation on the multiple historical surveillance videos respectively, combining every K adjacent surveillance images as a surveillance image group, and obtaining multiple historical surveillance image group sets, where K is a positive integer; Inputting each historical monitoring image group in the plurality of historical monitoring image group sets into an abnormal person identification branch and an abnormal event identification branch in a pre-trained abnormality identifier, respectively, and identifying and outputting abnormal persons and abnormal events in each historical monitoring image group; According to the time sequence of the historical monitoring image groups in each historical monitoring image group set, the identified abnormal persons and abnormal events are arranged to obtain multiple abnormal person sequences and multiple abnormal event sequences.

4. The monitoring resource scheduling method of the intelligent video monitoring system according to claim 3, characterized in that: The pre-training step of the anomaly identifier includes: Convolutional neural network is used to construct abnormal personnel identification branch and abnormal event identification branch respectively; According to the monitoring data records, a set of sample monitoring image groups is collected, and abnormal persons and abnormal events in each sample monitoring image group are marked to obtain a set of sample abnormal persons and a set of sample abnormal events; The sample monitoring image group set is used as input features, and the sample abnormal person set and the sample abnormal event set are used as output features, respectively, and supervised training and testing are performed on the abnormal person identification branch and the abnormal event identification branch, respectively, until the accuracy rate meets the convergence threshold; The converged abnormal person identification branch and abnormal event identification branch are combined to obtain an abnormal identifier.

5. The monitoring resource scheduling method of the intelligent video monitoring system according to claim 1, characterized in that: Performing personnel continuous abnormality analysis according to the multiple abnormal personnel sequences to obtain multiple personnel continuous abnormality degrees includes: Screening duplicate abnormal persons in the multiple abnormal person sequences to obtain multiple numbers of duplicate persons; Obtaining the number of abnormal persons in the plurality of abnormal person sequences to obtain a plurality of numbers of abnormal persons; The ratios of the number of repeated persons to the number of abnormal persons are calculated respectively to obtain the continuous abnormality degrees of the multiple persons.

6. The monitoring resource scheduling method of the intelligent video monitoring system according to claim 1, characterized in that: According to the multiple abnormal event sequences, event continuity abnormality analysis is performed according to the event chain library to obtain multiple event continuity abnormality degrees, including: In the plurality of abnormal event sequences, N abnormal events following each abnormal event are respectively combined to obtain a plurality of abnormal event group sets, where N is a positive integer; Input the plurality of abnormal event groups into an event chain library for retrieval to obtain a plurality of event chain quantities, wherein the event chain library includes a plurality of sample abnormal event chains, and when the abnormal event group is consistent with any one of the sample abnormal event chains, the event chain quantity is counted as 1; The ratios of the number of the multiple event chains and the number of abnormal event groups in the multiple abnormal event group sets are calculated respectively to obtain the continuous abnormality degrees of the multiple events.

7. The monitoring resource scheduling method of the intelligent video monitoring system according to claim 1, characterized in that: According to the multiple abnormal personnel sequences and the multiple abnormal event sequences, an abnormal density analysis is performed to obtain multiple abnormal densities, including: Calculating the total number of abnormal persons and the total number of abnormal events in the multiple abnormal person sequences and the multiple abnormal event sequences to obtain multiple numbers of abnormal persons and multiple numbers of abnormal events; Calculate the ratio of each abnormal person number to the average of multiple abnormal person numbers to obtain multiple abnormal person densities; Calculate the ratio of each abnormal event number to the average of multiple abnormal event numbers to obtain the abnormal density of multiple events; According to multiple personnel abnormal densities and multiple event abnormal densities, multiple abnormal densities are calculated to obtain multiple abnormal densities.

8. The monitoring resource scheduling method of the intelligent video monitoring system according to claim 1, characterized in that: According to the multiple personnel continuous abnormality degrees, the multiple event continuous abnormality degrees and the multiple abnormality densities, the multiple monitoring devices are allocated and scheduled for storage resources in the video storage server to obtain multiple storage resources as monitoring resource scheduling results, including: According to the plurality of personnel continuous abnormality degrees, the plurality of event continuous abnormality degrees and the plurality of abnormal densities, respectively, a plurality of first allocation coefficients, a plurality of second allocation coefficients and a plurality of third allocation coefficients are obtained by allocation calculation; Calculate and obtain a plurality of allocation scheduling coefficients according to the plurality of first allocation coefficients, the plurality of second allocation coefficients and the plurality of third allocation coefficients; According to the multiple allocation scheduling coefficients, the storage resources in the video storage server are allocated and scheduled to obtain multiple storage resources, and the monitoring video storage update records are performed on the multiple monitoring devices as the monitoring resource scheduling result.

9. A monitoring resource scheduling system for an intelligent video monitoring system, characterized in that: A monitoring resource scheduling method for an intelligent video monitoring system for implementing any one of claims 1 to 8, the system comprising: A video retrieval module, used to retrieve multiple historical surveillance videos of multiple surveillance devices through a video storage server in the video surveillance system, wherein the multiple surveillance devices are connected to the video storage server and send the surveillance videos to the video storage server for updating and storage; An abnormality identification module, used to identify abnormal persons and abnormal events according to the multiple historical surveillance videos, and obtain multiple abnormal person sequences and multiple abnormal event sequences; An abnormality analysis module is used to perform personnel continuous abnormality analysis according to the multiple abnormal personnel sequences to obtain multiple personnel continuous abnormality degrees, perform event continuous abnormality analysis according to the multiple abnormal event sequences and the event chain library to obtain multiple event continuous abnormality degrees, and perform abnormal density analysis according to the multiple abnormal personnel sequences and the multiple abnormal event sequences to obtain multiple abnormal densities; The allocation and scheduling module is used to allocate and schedule the storage resources of the multiple monitoring devices in the video storage server according to the multiple continuous abnormality degrees of the multiple personnel, the multiple continuous abnormality degrees of the multiple events and the multiple abnormality densities, and obtain multiple storage resources as the monitoring resource scheduling result.

Images