Edge computing optimization method and system for intelligent video surveillance system

Through edge computing, optimize the intelligent video surveillance system, dynamic packets and intelligent filter video streams, analyze and mark video frames in real time, generate hierarchical feedback instructions, optimize edge computing node resources, solve the network bandwidth pressure and delay problems in traditional systems, and improve monitoring System performance and response speed.

CN118552882BActive Publication Date: 2024-11-15SHENZHEN XUJING DIGITAL TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411014206.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-26
Publication Date
2024-11-15
Estimated Expiration
2044-07-26

AI Technical Summary

Technical Problem

Traditional intelligent video surveillance systems rely on central servers for data processing, resulting in high network bandwidth pressure, data transmission delay and packet loss, affecting the real-time and accuracy of monitoring.

Method used

Adopting edge computing optimization method, dynamic grouping and intelligent filtering by obtaining multi-angle video streams, analyzing video frames in real time, building a motion behavior system, marking video frames, and generating hierarchical feedback instructions, optimizing edge computing node resources, and improving the performance of monitoring system and response speed.

Benefits of technology

Enhanced coverage and depth of the monitoring system, improve the accuracy of abnormal behavior detection and continuity of specific event tracking, improve the real-time and responsiveness of the monitoring system, reduce data transmission delay, improve the efficiency of the overall system and Safety protection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118552882B_ABST
    Figure CN118552882B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of video optimization, and proposes an edge computing optimization method and system for an intelligent video monitoring system, the method comprising: obtaining multi-angle video streams and dynamically grouping them to obtain dynamic video groups, identifying video frame sequences for intelligent filtering to obtain filtered video frames, analyzing them to obtain video analysis events, generating discrete symbols and symbol strings to construct a motion behavior system and calculate behavior deviation values, marking, rendering and enhancing videos based on deviation values, transmitting them to a preset cloud center to obtain a risk assessment index for evaluation, determining importance grading to generate graded feedback instructions, executing instructions to obtain execution feedback data, optimizing node resources to obtain edge optimization parameters, and generating edge optimization reports. The present invention can improve the performance, reliability and response speed of the monitoring system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of video optimization, and in particular to an edge computing optimization method and system for an intelligent video surveillance system. Background Art

[0002] In today's information society, intelligent video surveillance systems play a vital role in ensuring public safety, corporate management and personal life. Especially in the context of the rapid development of big data and artificial intelligence technology, the requirements for video surveillance are becoming increasingly higher. Not only high-definition image quality and real-time monitoring are required, but also powerful data analysis and processing capabilities are required.

[0003] At present, traditional intelligent video surveillance systems mainly rely on central servers for data processing and analysis. Although this centralized processing method can meet basic monitoring needs, it has significant limitations. For example, the transmission of a large amount of video data to the central server will cause huge pressure on the network bandwidth, which is easy to cause data transmission delays and packet loss, affecting the real-time and accuracy of monitoring. Therefore, an edge computing optimization method for intelligent video surveillance systems is needed to improve the performance, reliability and response speed of the monitoring system. Summary of the invention

[0004] The present invention provides an edge computing optimization method and system for an intelligent video surveillance system, the main purpose of which is to improve the performance, reliability and response speed of the surveillance system.

[0005] Acquire a multi-angle video stream in an intelligent video surveillance system, dynamically group the multi-angle video stream to obtain a dynamic video group, identify a video frame sequence corresponding to the dynamic video group, and intelligently filter the dynamic video group based on the video frame sequence to obtain a filtered video frame;

[0006] Performing real-time analysis on the filtered video frame to obtain video analysis events, wherein the video analysis events include video abnormal behavior events and specific object tracking events, generating discrete symbols and symbol strings corresponding to the video analysis events, constructing a motion behavior system corresponding to the filtered video frame based on the discrete symbols and the symbol strings, and calculating a behavior deviation value corresponding to the motion behavior system and a preset standard behavior pattern;

[0007] Based on the behavior deviation value, the filtered video frame is marked to obtain a marked video frame, the marked video frame is quickly rendered to obtain a rendered video, and the rendered video is enhanced to obtain an enhanced video;

[0008] Transmitting the enhanced video to a preset cloud center for risk assessment to obtain a risk assessment index, determining the importance level of the enhanced video based on the risk assessment index, and generating a graded feedback instruction corresponding to the multi-angle video stream based on the importance level;

[0009] After executing the hierarchical feedback instruction using the preset edge computing node, execution feedback data is obtained. Based on the execution feedback data, edge optimization is performed on the node resources in the preset edge computing node to obtain edge optimization parameters. Based on the edge optimization parameters, an edge optimization report corresponding to the intelligent video surveillance system is generated.

[0010] Optionally, dynamically grouping the multi-angle video streams to obtain a dynamic video group includes:

[0011] Performing unified format conversion on the multi-angle video streams to obtain a unified video stream;

[0012] Performing scene recognition on the unified video stream to obtain a video scene;

[0013] Extracting dynamic features in the video scene;

[0014] Based on the dynamic features, the multi-angle video streams are dynamically grouped to obtain dynamic video groups.

[0015] Optionally, the intelligently filtering the dynamic video group based on the video frame sequence to obtain filtered video frames includes:

[0016] Identify a sequence number corresponding to the video frame sequence;

[0017] Based on the sequence number, adjacent frames in the dynamic video group are determined, and adjacent frame differences corresponding to the adjacent frames are calculated;

[0018] Based on the adjacent frame differences, determining a dynamic region corresponding to the dynamic video group, and querying a filter point in the dynamic region;

[0019] Based on the filtering points, the dynamic video group is intelligently filtered to obtain filtered video frames.

[0020] Optionally, the calculating adjacent frame differences corresponding to the adjacent frames includes:

[0021] The adjacent frame differences corresponding to the adjacent frames are calculated using the following formula:

[0022] ;

[0023] in, Indicates Frame and The difference between adjacent frames of a frame, represents the row index of the image corresponding to the adjacent frame, represents the number of rows of the corresponding image of the adjacent frames, represents the column index of the image corresponding to the adjacent frame, represents the number of columns of the corresponding images of the adjacent frames, Indicates Frame at coordinates The pixel value at Indicates Frame at coordinates The pixel value at Representing coordinates The weight coefficient of the pixel at , α represents the pixel sensitivity of the corresponding image of the adjacent frame.

[0024] Optionally, constructing a motion behavior system corresponding to the filtered video frame based on the discrete symbol and the symbol string includes:

[0025] Classify the discrete symbol and the symbol string to obtain a classification symbol;

[0026] Extracting key features of symbols from the classified symbols;

[0027] Initially constructing behavior patterns corresponding to the key features of the symbols, and analyzing correlations between the behavior patterns;

[0028] Based on the association relationship, a motion behavior system corresponding to the filtered video frame is constructed.

[0029] Optionally, the calculating of the behavior deviation value between the movement behavior system and a preset standard behavior pattern includes:

[0030] The behavior deviation value corresponding to the sports behavior system and the preset standard behavior pattern is calculated using the following formula:

[0031] ;

[0032] in, represents the behavior deviation value corresponding to the sports behavior system and the preset standard behavior pattern, An index representing the behavior characteristics corresponding to the sports behavior system and the preset standard behavior pattern, Indicates the number of behavioral features corresponding to the sports behavior system and the preset standard behavior pattern, Indicates The weight of the behavioral characteristics, Indicates the first The actual value of the behavior characteristic, Indicates the first The standard value of a behavioral characteristic.

[0033] Optionally, the step of marking the filtered video frame based on the behavior deviation value to obtain a marked video frame includes:

[0034] Based on a preset threshold range, the behavior deviation value is classified to obtain a classification deviation value;

[0035] Identifying the abnormal behavior type corresponding to the classification deviation value;

[0036] Based on the abnormal behavior type, generating a marking strategy corresponding to the filtered video frame;

[0037] A video frame region corresponding to the marking strategy is determined, and the video frame region is marked to obtain a marked video frame.

[0038] Optionally, the generating, based on the importance classification, a graded feedback instruction corresponding to the multi-angle video stream includes:

[0039] Determine a classification label corresponding to the importance classification;

[0040] Formulate a feedback strategy corresponding to the multi-angle video stream based on the classification label;

[0041] Querying feedback parameters in the feedback strategy;

[0042] Based on the feedback parameters, generating a primary feedback instruction corresponding to the multi-angle video stream;

[0043] Performing simulated execution on the primary feedback instruction to obtain simulated execution data;

[0044] Based on the simulation execution data, the primary feedback instructions are hierarchically optimized to obtain hierarchical feedback instructions.

[0045] Optionally, after executing the hierarchical feedback instruction by using a preset edge computing node, execution feedback data is obtained, including:

[0046] Parsing the hierarchical feedback instruction to obtain parsed instruction data;

[0047] Querying the execution environment corresponding to the parsed instruction data;

[0048] Identifying execution tasks in the execution environment; prioritizing the execution tasks to obtain task priorities;

[0049] Based on the task priority, a preset edge computing node is used to perform feedback analysis on the execution task to obtain execution feedback data.

[0050] In order to solve the above problems, the present invention also provides an edge computing optimization system for an intelligent video surveillance system, the system comprising:

[0051] An intelligent filtering module is used to obtain a multi-angle video stream in an intelligent video surveillance system, dynamically group the multi-angle video stream to obtain a dynamic video group, identify a video frame sequence corresponding to the dynamic video group, and intelligently filter the dynamic video group based on the video frame sequence to obtain a filtered video frame;

[0052] A deviation value calculation module is used to perform real-time analysis on the filtered video frame to obtain video analysis events, wherein the video analysis events include video abnormal behavior events and specific object tracking events, generate discrete symbols and symbol strings corresponding to the video analysis events, construct a motion behavior system corresponding to the filtered video frame based on the discrete symbols and the symbol strings, and calculate a behavior deviation value corresponding to the motion behavior system and a preset standard behavior pattern;

[0053] A video enhancement module, used for marking the filtered video frame based on the behavior deviation value to obtain a marked video frame, quickly rendering the marked video frame to obtain a rendered video, and performing video enhancement on the rendered video to obtain an enhanced video;

[0054] An instruction generation module is used to transmit the enhanced video to a preset cloud center for risk assessment, obtain a risk assessment index, determine the importance level corresponding to the enhanced video based on the risk assessment index, and generate a graded feedback instruction corresponding to the multi-angle video stream based on the importance level;

[0055] A report generation module is used to obtain execution feedback data after executing the hierarchical feedback instruction using a preset edge computing node, and based on the execution feedback data, edge optimize the node resources in the preset edge computing node to obtain edge optimization parameters, and based on the edge optimization parameters, generate an edge optimization report corresponding to the intelligent video surveillance system.

[0056] First, the present invention greatly enhances the coverage and depth of the monitoring system by acquiring multi-angle video streams in the intelligent video monitoring system, making the detection of abnormal behavior more accurate and the tracking of specific events more continuous, thereby optimizing video management and improving the efficiency of the monitoring system. At the same time, the present invention obtains video analysis events by performing real-time analysis on the filtered video frames, greatly enhancing the real-time and responsiveness of the monitoring system, making the detection of abnormal behavior more rapid and the tracking of specific events more continuous, thereby improving security protection and optimizing monitoring strategies. The present invention marks the filtered video frames based on the behavior deviation values ​​to obtain marked video frames, which helps to quickly locate abnormal behaviors or situations that do not meet expectations. And identification, can quickly screen out the parts that need to be focused on, improve the efficiency of monitoring and analysis, the present invention transmits the enhanced video to the preset cloud center for risk assessment, obtains the risk assessment index, can accurately identify and quantify various potential risks in the video, and can comprehensively, deeply and efficiently analyze the complex enhanced video, so as to obtain a more accurate and reliable risk assessment index. Furthermore, the present invention obtains execution feedback data after executing the hierarchical feedback instruction by using the preset edge computing node, which helps to improve the response speed and efficiency of the system, can reduce the delay of data transmission, enable the execution feedback data to be acquired and analyzed faster, respond quickly, and improve the operation efficiency of the entire system. Therefore, the edge computing optimization method and system of the intelligent video monitoring system proposed by the present invention can improve the performance, reliability and response speed of the monitoring system. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A schematic diagram of a flow chart of an edge computing optimization method for an intelligent video surveillance system provided by an embodiment of the present invention;

[0058] Figure 2 A module diagram of an edge computing optimization system for an intelligent video surveillance system provided by one embodiment of the present invention.

[0059] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0060] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0061] The embodiment of the present application provides an edge computing optimization method for an intelligent video surveillance system. The execution subject of the edge computing optimization method of the intelligent video surveillance system includes but is not limited to at least one of the electronic devices such as a server, a terminal, etc. that can be configured to execute the method provided by the embodiment of the present application. In other words, the edge computing optimization method of the intelligent video surveillance system can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms.

[0062] Reference Figure 1 FIG. 1 is a flow chart of an edge computing optimization method for an intelligent video surveillance system provided by an embodiment of the present invention. In this embodiment, the edge computing optimization method for the intelligent video surveillance system includes:

[0063] S1. Acquire a multi-angle video stream in an intelligent video surveillance system, dynamically group the multi-angle video stream to obtain a dynamic video group, identify a video frame sequence corresponding to the dynamic video group, and based on the video frame sequence, intelligently filter the dynamic video group to obtain a filtered video frame.

[0064] The present invention greatly enhances the coverage and depth of the monitoring system by acquiring multi-angle video streams in the intelligent video monitoring system, making the detection of abnormal behavior more accurate and the tracking of specific events more continuous, thereby optimizing video management and improving the efficiency of the monitoring system.

[0065] Among them, the intelligent video surveillance system refers to an advanced monitoring system that integrates advanced technologies such as artificial intelligence, machine learning, image recognition, etc., which can automatically analyze and process video data, identify abnormal behavior, track specific targets, and provide real-time alarms and data reports; the multi-angle video stream refers to the video data stream captured simultaneously by cameras installed at multiple different positions and angles. These video streams can be shot from different directions, different heights or different distances, which together constitute all-round coverage of the monitored area. Optionally, the multi-angle video stream in the intelligent video surveillance system can be obtained through a camera network, such as: using multiple high-definition cameras installed at different positions and angles.

[0066] Furthermore, the present invention dynamically groups the multi-angle video streams to obtain dynamic video groups, which can flexibly adjust the monitoring focus according to the changes in the real-time scene, and can also effectively integrate information from different perspectives to improve the efficiency and accuracy of data processing.

[0067] The dynamic video group refers to a combination formed by grouping multi-angle video streams with similar or related dynamic features together based on the extracted dynamic features. For example, in the monitoring of a traffic intersection, the video stream of fast-moving vehicles may be divided into a dynamic video group, while the video stream of slowly walking pedestrians may be divided into another group.

[0068] As an embodiment of the present invention, the dynamic grouping of the multi-angle video streams to obtain dynamic video groups includes: converting the multi-angle video streams into a unified format to obtain a unified video stream; performing scene recognition on the unified video stream to obtain a video scene; extracting dynamic features in the video scene; and dynamically grouping the multi-angle video streams based on the dynamic features to obtain a dynamic video group.

[0069] The unified video stream refers to converting multi-angle video streams into the same format to eliminate format differences between videos from different sources, so as to facilitate subsequent unified processing and analysis. For example, videos from cameras of different models are converted into video streams in H.264 encoding format; the video scene refers to the specific environment and situation presented in the video, including a combination of elements such as location, time, people, and objects, such as a work scene in an office, a shopping scene in a mall, or a classroom scene in a school; the dynamic features refer to features in the video scene that can reflect the motion state and changes of objects or elements, such as the moving speed, direction, acceleration of an object, or the amplitude and frequency of a person's movements.

[0070] Furthermore, the unified format conversion of the multi-angle video stream can be achieved through a video encoder, such as FFmpeg and other tools; the scene recognition of the unified video stream can be achieved through a pre-trained model, such as VGGNet, ResNet and other models; the extraction of dynamic features in the video scene can be achieved through a feature extraction algorithm, such as SIFT, SURF and other algorithms; the dynamic grouping of the multi-angle video stream can be achieved through a clustering grouping algorithm, such as K-means, DBSCAN and other algorithms.

[0071] By identifying the video frame sequence corresponding to the dynamic video group, the present invention can enhance the tracing ability and analysis depth of the monitoring system, making the detection of abnormal behavior more accurate and the tracking of specific events more continuous, thereby improving security protection and optimizing monitoring strategies.

[0072] Among them, the video frame sequence refers to a set of continuous video frames arranged in a certain order within a period of time, each frame is a static image in the video, and these frames are played in chronological order to form a dynamic video effect. Optionally, the video frame sequence corresponding to the dynamic video group can be obtained by real-time stream processing platform tools, such as: Apache Kafka, Apache Flink and other tools.

[0073] Furthermore, the present invention intelligently filters the dynamic video group based on the video frame sequence to obtain filtered video frames, which greatly enhances the focus and insight of the monitoring system, makes the detection of abnormal behavior more accurate, and the tracking of specific events more continuous, thereby greatly improving the efficiency and accuracy of data processing.

[0074] The filtered video frame refers to a video frame obtained after intelligent filtering processing, from which parts that are determined to not meet specific conditions or have no important information are removed.

[0075] As an embodiment of the present invention, the dynamic video group is intelligently filtered based on the video frame sequence to obtain filtered video frames, including: identifying a serial number corresponding to the video frame sequence; determining adjacent frames in the dynamic video group based on the serial number; calculating adjacent frame differences corresponding to the adjacent frames; determining a dynamic area corresponding to the dynamic video group based on the adjacent frame differences; querying filtering points in the dynamic area; and intelligently filtering the dynamic video group based on the filtering points to obtain filtered video frames.

[0076] Among them, the serial number refers to a unique identifier assigned to each frame in the video frame sequence, which is used to identify and distinguish different frames, and is usually incremented in chronological order; the adjacent frames refer to two frames that are close to each other in sequence in the video frame sequence, such as the nth frame and the n+1th frame; the adjacent frame difference refers to a difference measure of pixel values ​​between two adjacent frames, which is obtained by calculating the difference in pixel values ​​of corresponding pixel positions in adjacent frames, and reflects the degree of change of image content between adjacent frames; the dynamic area refers to an area in the video where there is object movement or scene change, which is determined by analyzing the adjacent frame difference, and is usually manifested as a part with a large change in pixel value; the filter point refers to a key position or pixel point in the dynamic area that is identified as requiring a filtering operation, and its characteristics may include abnormal pixel values, non-compliance with preset rules, etc.

[0077] Furthermore, the identification of the serial number corresponding to the video frame sequence can be obtained through a sequence generation model, such as RNN, LSTM and other models; the determination of adjacent frames in the dynamic video group can be obtained through a frame extraction algorithm, such as frame difference method and other algorithms; the calculation of adjacent frame differences corresponding to the adjacent frames can be obtained through the following algorithm, the determination of the dynamic area corresponding to the dynamic video group can be obtained through a motion detection algorithm, such as optical flow method, background subtraction method and other algorithms; the query of filter points in the dynamic area can be obtained through a target detection model, such as YOLO, SSD model; the intelligent filtering of the dynamic video group can be obtained through a filtering model, such as Caffe framework and other models.

[0078] As an embodiment of the present invention, the calculating adjacent frame differences corresponding to the adjacent frames includes:

[0079] The adjacent frame differences corresponding to the adjacent frames are calculated using the following formula:

[0080] ;

[0081] in, Indicates Frame and The difference between adjacent frames of a frame, represents the row index of the image corresponding to the adjacent frame, represents the number of rows of the corresponding image of the adjacent frames, represents the column index of the image corresponding to the adjacent frame, represents the number of columns of the corresponding images of the adjacent frames, Indicates Frame at coordinates The pixel value at Indicates Frame at coordinates The pixel value at Representing coordinates The weight coefficient of the pixel at , α represents the pixel sensitivity of the corresponding image of the adjacent frame.

[0082] S2. Perform real-time analysis on the filtered video frames to obtain video analysis events, which include abnormal video behavior events and specific object tracking events. Generate discrete symbols and symbol strings corresponding to the video analysis events, construct a motion behavior system corresponding to the filtered video frames based on the discrete symbols and the symbol strings, and calculate the behavior deviation value corresponding to the motion behavior system and the preset standard behavior pattern.

[0083] The present invention obtains video analysis events by performing real-time analysis on the filtered video frames, thereby greatly enhancing the real-time performance and response capability of the monitoring system, making the detection of abnormal behaviors more rapid and the tracking of specific events more continuous, thereby improving security protection and optimizing monitoring strategies.

[0084] Among them, the video analysis event refers to a specific event or situation identified after analyzing and processing the video content. These events usually involve dynamic changes in the monitoring scene, such as human behavior, object movement, the occurrence of abnormal situations, etc. The video analysis event includes video abnormal behavior events and specific object tracking events. The abnormal behavior event refers to abnormal or unexpected behavior that occurs in the monitoring scene, such as intrusion, wandering, violent behavior, etc.; specific object tracking events refer to the system's continuous tracking and monitoring of specific objects (such as vehicles, pedestrians, abandoned items, etc.), recording their movement paths and behavior patterns.

[0085] Furthermore, by generating discrete symbols and symbol strings corresponding to the video analysis events, the present invention can transmit more efficiently, reduce the amount of data and transmission time, improve the overall performance of the system, enable the system to respond more quickly and accurately, and enhance the intelligence and practicality of the entire video surveillance system.

[0086] Among them, the discrete symbol refers to an independent symbol used to individually represent a specific meaning or attribute under specific rules or definitions. For example, in video analysis, "A" is used to represent "person appearance" and "B" represents "vehicle movement". Here, "A" and "B" are discrete symbols; the symbol string refers to a symbol sequence composed of multiple discrete symbols arranged and combined in a certain order. For example, in the above example, "AB" can represent "person appearance and vehicle movement", which is a symbol string. Optionally, the generation of discrete symbols and symbol strings corresponding to the video analysis events can be obtained through character generation algorithms, such as: hash algorithms, encoding algorithms and other algorithms.

[0087] Furthermore, the present invention constructs a motion behavior system corresponding to the filtered video frame based on the discrete symbols and the symbol string, which can not only accurately describe and record the characteristics of key behaviors, but also reveal the internal logic and development context of the behavior through symbolic representation, thereby enhancing the analysis ability and understanding depth of the monitoring system.

[0088] Among them, the motion behavior system refers to a comprehensive and complete system, which includes various different behavior patterns and the relationships between them, and can comprehensively and systematically describe the various motion behaviors and their relationships presented in the filtered video frames. For example, it covers a complete set of descriptions and rules for various movement methods of characters, state changes of objects, and the interactions between them.

[0089] As an embodiment of the present invention, the motion behavior system corresponding to the filtered video frame is constructed based on the discrete symbols and the symbol string, including: classifying the discrete symbols and the symbol string to obtain classification symbols; extracting the symbol key features in the classification symbols; initially constructing the behavior pattern corresponding to the symbol key features; analyzing the correlation between the behavior patterns; and constructing the motion behavior system corresponding to the filtered video frame based on the correlation.

[0090] Among them, the classification symbol refers to a set of symbols with the same or similar attributes obtained by classifying and arranging discrete symbols and symbol strings according to certain rules or standards. For example, symbols "P1", "P2" and the like representing rapid movement of people are classified into one category, and symbols "O1", "O2" and the like representing stationary objects are classified into another category. Each category here is a classification symbol; the key features of the symbol refer to the features extracted from the classification symbol that can reflect its essence and important attributes. For example, for the classification symbol representing rapid movement of people, the key features may include movement speed, direction, duration, etc.; the behavior pattern refers to a behavioral expression with certain regularity and repeatability composed of the key features of the symbol. For example, "a person moves continuously in the southeast direction at a speed of 5 meters per second for 10 seconds" is a behavior pattern; the association relationship refers to the relationship of mutual connection, influence or dependence between different behavior patterns. For example, the behavior pattern of "a person moves quickly in a certain direction" may be associated with the behavior pattern of "there is an object that attracts the person in that direction".

[0091] Furthermore, the symbol classification of the discrete symbols and the symbol strings can be achieved through NLP tools, such as NLTK, spaCy and other tools; the extraction of key symbol features in the classified symbols can be achieved through feature extraction algorithms, such as PCA, ICA and other algorithms; the initial construction of the behavior patterns corresponding to the key symbol features can be achieved through machine learning models, such as support vector machines, random forests and other models; the analysis of the association between the behavior patterns can be achieved through association rule learning algorithms, such as Apriori algorithms, FP-Growth algorithms and other algorithms; the construction of the motion behavior system corresponding to the filtered video frames can be achieved through graph construction tools, such as Neo4j, GraphDB and other tools.

[0092] By calculating the behavior deviation value corresponding to the motion behavior system and the preset standard behavior pattern, the present invention can take measures to intervene in advance to avoid further deterioration of the problem, and can allocate resources in a targeted manner to solve the problem or improve the system to improve resource utilization efficiency.

[0093] The behavior deviation value refers to the quantitative value of the difference between the actual behavior characteristics presented by the sports behavior system and the preset standard behavior pattern.

[0094] As an embodiment of the present invention, the calculating of the behavior deviation value corresponding to the sports behavior system and the preset standard behavior pattern includes:

[0095] The behavior deviation value corresponding to the sports behavior system and the preset standard behavior pattern is calculated using the following formula:

[0096] ;

[0097] in, represents the behavior deviation value corresponding to the sports behavior system and the preset standard behavior pattern, An index representing the behavior characteristics corresponding to the sports behavior system and the preset standard behavior pattern, Indicates the number of behavioral features corresponding to the sports behavior system and the preset standard behavior pattern, Indicates The weight of the behavioral characteristics, Indicates the first The actual value of the behavior characteristic, Indicates the first The standard value of a behavioral characteristic.

[0098] S3. Based on the behavior deviation value, mark the filtered video frame to obtain a marked video frame, quickly render the marked video frame to obtain a rendered video, and enhance the rendered video to obtain an enhanced video.

[0099] The present invention marks the filtered video frames based on the behavior deviation values ​​to obtain marked video frames, which helps to quickly locate and identify abnormal behaviors or situations that do not meet expectations, and can quickly screen out parts that need to be focused on, thereby improving the efficiency of monitoring and analysis.

[0100] The marked video frame refers to a video frame with a specific mark obtained after a marking operation, which is used to intuitively display and distinguish different abnormal behaviors or deviation situations.

[0101] As an embodiment of the present invention, the filtered video frame is marked based on the behavior deviation value to obtain a marked video frame, including: based on a preset threshold range, the behavior deviation value is classified to obtain a classification deviation value; identifying the abnormal behavior type corresponding to the classification deviation value; based on the abnormal behavior type, generating a marking strategy corresponding to the filtered video frame; determining the video frame area corresponding to the marking strategy; marking the video frame area to obtain a marked video frame.

[0102] Among them, the classification deviation value refers to the different categories obtained after dividing the behavior deviation value according to a preset threshold range, for example, the deviation value is divided into high deviation, medium deviation and low deviation categories; the abnormal behavior type refers to the specific behavior category that is obviously different from the preset standard behavior pattern identified by analyzing the classification deviation value, such as speeding behavior, abnormal steering behavior, etc.; the marking strategy refers to the specific method and rules for marking filtered video frames based on the abnormal behavior type, for example, using specific colors, shapes or symbols to mark different abnormal behaviors; the video frame area refers to the specific part that needs to be marked in the filtered video frame, which may be an object or a specific position range in the entire frame.

[0103] Furthermore, the classification of the behavior deviation values ​​can be achieved through a threshold classification algorithm, such as a fixed threshold method, an adaptive threshold method, and other algorithms; the identification of the abnormal behavior type corresponding to the classification deviation value can be achieved through a deep learning model, such as CNN, RNN; the generation of the marking strategy corresponding to the filtered video frame can be achieved through an NLP tool, such as NLTK, spaCy, and other tools; the determination of the video frame area corresponding to the marking strategy can be achieved through a target detection model, such as YOLO, SSD, and other models; the marking of the video frame area can be achieved through an image processing library tool, such as OpenCV and other tools.

[0104] The present invention obtains a rendered video by quickly rendering the marked video frame, which can greatly improve the visual effect and quality of the video, make the mark clearer and more eye-catching, and enhance the communication and display effect of information.

[0105] Among them, the rendered video refers to a video that has been rendered and optimized. These videos usually contain markings, enhancements or other visual effects to more clearly display key information in the monitoring scene. Optionally, the fast rendering of the marked video frames can be achieved through a graphics rendering engine, such as: OpenGL, DirectX and other tools.

[0106] Furthermore, the present invention performs video enhancement on the rendered video to obtain an enhanced video, which can significantly improve the clarity and detail of the image, so that important information hidden in the video can be presented more clearly, which is conducive to more accurate analysis and judgment, and can improve the contrast and brightness of the video, making the picture more vivid and layered, and can restore more realistic and accurate colors.

[0107] Among them, the enhanced video refers to a video that has been significantly improved and enhanced in terms of clarity, contrast, brightness, color, details, etc. of the original video through a series of image processing and optimization technologies. Optionally, the video enhancement of the rendered video can be achieved through an image enhancement algorithm, such as: Retinex algorithm, bilateral filtering and other algorithms.

[0108] S4. Transmit the enhanced video to a preset cloud center for risk assessment to obtain a risk assessment index, determine the importance grade corresponding to the enhanced video based on the risk assessment index, and generate a graded feedback instruction corresponding to the multi-angle video stream based on the importance grade.

[0109] The present invention transmits the enhanced video to a preset cloud center for risk assessment to obtain a risk assessment index, which can accurately identify and quantify various potential risks in the video, and can perform a comprehensive, in-depth and efficient analysis of complex enhanced videos, thereby obtaining a more accurate and reliable risk assessment index.

[0110] Among them, the preset cloud center refers to a pre-set and configured cloud computing service center, which has powerful computing power, massive storage space and perfect network connection, and can centrally process and analyze large amounts of data and complex tasks; the risk assessment index refers to a quantitative value obtained through a series of risk assessment processes and calculations on a specific object (such as the situation reflected in the enhanced video). Optionally, the enhanced video is transmitted to the preset cloud center for risk assessment, which can be achieved through network security tools, such as: VPN, SSL, TLS and other tools.

[0111] Furthermore, the present invention determines the importance classification corresponding to the enhanced video based on the risk assessment index, and can quickly focus on those videos that have a greater impact on business or security, ensuring that key issues are handled in a timely manner, and can provide a clear basis for subsequent decision-making.

[0112] Among them, the importance grading refers to dividing things (enhanced videos in this context) into different levels or layers according to their importance based on specific standards and evaluation indicators. Optionally, the importance grading corresponding to the enhanced video can be obtained through data analysis tools, such as: Tableau, PowerBI and other tools.

[0113] The present invention generates graded feedback instructions corresponding to the multi-angle video stream based on the importance classification, and can timely allocate more bandwidth, storage and computing resources to ensure smooth transmission and efficient processing, avoid affecting the acquisition of key information due to insufficient resources, and clearly define the priority and urgency of processing, so as to quickly take corresponding actions and reduce decision-making time and the possibility of misjudgment.

[0114] The hierarchical feedback instructions refer to precise processing instructions for multi-angle video streams adapted to different importance levels, which are finally determined after simulating and optimizing the primary feedback instructions.

[0115] As an embodiment of the present invention, the generation of graded feedback instructions corresponding to the multi-angle video stream based on the importance grading includes: determining a grading label corresponding to the importance grading; formulating a feedback strategy corresponding to the multi-angle video stream based on the grading label; querying feedback parameters in the feedback strategy; generating primary feedback instructions corresponding to the multi-angle video stream based on the feedback parameters; simulating the execution of the primary feedback instructions to obtain simulated execution data; and performing graded optimization on the primary feedback instructions based on the simulated execution data to obtain graded feedback instructions.

[0116] Among them, the grading label refers to a specific identifier or symbol set for the multi-angle video stream according to the importance classification, which is used to distinguish video streams of different importance; the feedback strategy refers to; the feedback parameter refers to the numerical value or indicator specifically specified in the feedback strategy, such as the length of the response time, the proportion of resource allocation, etc.; the primary feedback instruction refers to the processing instruction for the multi-angle video stream that is initially generated based on the feedback parameter and has not yet been optimized and improved; the simulated execution data refers to the data generated by the simulated operation of the primary feedback instruction, including various situations, effects, resource consumption and other information during the execution process.

[0117] Furthermore, the determination of the graded labels corresponding to the importance grades can be obtained through a threshold classification algorithm, such as a fixed threshold method, an adaptive threshold method and other algorithms; the formulation of the feedback strategy corresponding to the multi-angle video stream can be obtained through a rule engine; the query of the feedback parameters in the feedback strategy can be obtained through a database management system, such as MySQL, PostgreSQL, etc.; the generation of the primary feedback instructions corresponding to the multi-angle video stream can be obtained through an instruction generator; the simulation execution of the primary feedback instructions can be obtained through a simulator, such as Simulink, AnyLogic and other tools; the hierarchical optimization of the primary feedback instructions can be obtained through a machine learning model, such as a support vector machine, a random forest and other models.

[0118] S5. After executing the hierarchical feedback instruction using the preset edge computing node, execution feedback data is obtained. Based on the execution feedback data, edge optimization is performed on the node resources in the preset edge computing node to obtain edge optimization parameters. Based on the edge optimization parameters, an edge optimization report corresponding to the intelligent video surveillance system is generated.

[0119] The present invention obtains execution feedback data by utilizing preset edge computing nodes to execute the hierarchical feedback instructions, which helps to improve the response speed and efficiency of the system, can reduce the delay in data transmission, enable the execution feedback data to be acquired and analyzed more quickly, respond quickly, and improve the operating efficiency of the entire system.

[0120] Among them, the preset edge computing nodes refer to computing nodes that are pre-set in the system and located at the edge of the network. These nodes are usually close to the data source or user terminal and have certain computing, storage and communication capabilities. They can process and analyze data close to the source of data generation, reduce data transmission delays and bandwidth consumption, and improve data processing efficiency and real-time performance; the execution feedback data refers to various data about the task execution status that are fed back by the preset edge computing node after completing the execution task, including execution results, resource consumption, problems encountered, etc.

[0121] As an embodiment of the present invention, after executing the hierarchical feedback instruction using the preset edge computing node, execution feedback data is obtained, including: performing instruction parsing on the hierarchical feedback instruction to obtain parsed instruction data; querying the execution environment corresponding to the parsed instruction data; identifying the execution tasks in the execution environment; prioritizing the execution tasks to obtain task priorities; based on the task priorities, performing feedback analysis on the execution tasks using the preset edge computing nodes to obtain execution feedback data.

[0122] Among them, the parsed instruction data refers to the detailed and clear instruction information obtained after analyzing and disassembling the hierarchical feedback instructions, including specific operation requirements, parameter settings, etc.; the execution environment refers to the sum of various conditions and factors in which the preset edge computing node is in when performing a task, such as network conditions, hardware resources, software configuration, etc.; the execution task refers to the specific work content and operation that the preset edge computing node needs to complete in accordance with the hierarchical feedback instructions; the task priority refers to the order of execution of tasks based on factors such as the importance and urgency of the task.

[0123] Furthermore, the instruction parsing of the hierarchical feedback instructions can be achieved through a parsing algorithm, such as a syntax analysis algorithm, such as a recursive descent algorithm and other algorithms; the query of the execution environment corresponding to the parsed instruction data can be achieved through an environment monitoring tool, such as Zabbix and other tools; the identification of the execution tasks in the execution environment can be achieved through a task identification tool, such as Task Identifier and other tools; the prioritization of the execution tasks can be achieved through a priority sorting tool, such as Priority Queue and other tools; the feedback analysis of the execution tasks can be achieved by utilizing a preset edge computing node.

[0124] The present invention performs edge optimization on node resources in preset edge computing nodes based on the execution feedback data to obtain edge optimization parameters, which can accurately understand the usage of node resources in various tasks, thereby performing targeted optimization to avoid idleness and waste of resources, help improve the response speed and performance of the system, and be able to process tasks more quickly, thereby enhancing the stability and reliability of the system.

[0125] The edge optimization parameter refers to a specific value or setting that can be used to adjust and configure resources and is determined during the edge optimization process of node resources in a preset edge computing node.

[0126] As an embodiment of the present invention, the node resources in the preset edge computing node are edge optimized based on the execution feedback data to obtain edge optimization parameters, including: data aggregation of the execution feedback data to obtain aggregated data; extracting resource usage records in the aggregated data; identifying resource balancing points in the resource usage records; based on the resource balancing points, determining optimization targets corresponding to the node resources in the preset edge computing node; based on the optimization targets, edge optimization of the node resources in the preset edge computing node to obtain edge optimization parameters.

[0127] Among them, the summary data refers to the data set obtained after the execution feedback data is centralized, integrated and comprehensively processed, which contains feedback information from various aspects; the resource usage record refers to a detailed record of the usage of various resources (such as computing power, storage, bandwidth, etc.) in the preset edge computing node during the execution of the task; the resource balance point refers to the key point or state in the resource usage record that can make the allocation of various resources reach a relative balance and optimal state; the optimization goal refers to the specific goal to be achieved based on the analysis and judgment of the resource balance point in order to achieve more efficient resource utilization, improve system performance, etc.

[0128] Furthermore, the data aggregation of the execution feedback data can be achieved through data integration tools, such as Apache NiFi and other tools; the extraction of resource usage records in the aggregated data can be achieved through data extraction tools, such as Kettle and other tools; the identification of resource balance points in the resource usage records can be achieved through data analysis tools, such as RapidMiner and other tools; the determination of the optimization target corresponding to the node resources in the preset edge computing node can be achieved through a target programming algorithm, such as the simplex method, hierarchical sequence method and other algorithms; the edge optimization of the node resources in the preset edge computing node can be achieved through an optimization solver, such as CPLEX and other tools.

[0129] The present invention generates an edge optimization report corresponding to the intelligent video surveillance system based on the edge optimization parameters, which can fully understand the improvement of key indicators such as resource utilization efficiency and monitoring effect of the system after edge optimization, thereby accurately judging the effectiveness of the optimization measures.

[0130] Among them, the edge optimization report refers to a comprehensive document generated based on the edge optimization parameters of the intelligent video surveillance system, which records in detail the edge optimization process, the measures taken, the comparison of system performance before and after optimization, the changes in resource allocation and other key information. At the same time, it also includes an evaluation and analysis of the optimization effect, as well as suggestions for further improvements or adjustments that may be needed in the future. Optionally, the edge optimization report corresponding to the intelligent video surveillance system can be obtained through a report generation tool, such as Microsoft Power BI, Tableau and other tools.

[0131] First, the present invention greatly enhances the coverage and depth of the monitoring system by acquiring multi-angle video streams in the intelligent video monitoring system, making the detection of abnormal behavior more accurate and the tracking of specific events more continuous, thereby optimizing video management and improving the efficiency of the monitoring system. At the same time, the present invention obtains video analysis events by performing real-time analysis on the filtered video frames, greatly enhancing the real-time and responsiveness of the monitoring system, making the detection of abnormal behavior more rapid and the tracking of specific events more continuous, thereby improving security protection and optimizing monitoring strategies. The present invention marks the filtered video frames based on the behavior deviation values ​​to obtain marked video frames, which helps to quickly locate abnormal behaviors or situations that do not meet expectations. And identification, can quickly screen out the parts that need to be focused on, improve the efficiency of monitoring and analysis, the present invention transmits the enhanced video to the preset cloud center for risk assessment, obtains the risk assessment index, can accurately identify and quantify various potential risks in the video, and can comprehensively, deeply and efficiently analyze the complex enhanced video, so as to obtain a more accurate and reliable risk assessment index. Furthermore, the present invention obtains execution feedback data after executing the hierarchical feedback instruction by using the preset edge computing node, which helps to improve the response speed and efficiency of the system, can reduce the delay of data transmission, enable the execution feedback data to be acquired and analyzed faster, respond quickly, and improve the operation efficiency of the entire system. Therefore, the edge computing optimization method and system of the intelligent video monitoring system proposed by the present invention can improve the performance, reliability and response speed of the monitoring system.

[0132] like Figure 2 , is a functional module diagram of an edge computing optimization system of an intelligent video surveillance system provided by one embodiment of the present invention.

[0133] The edge computing optimization system 200 of the intelligent video surveillance system of the present invention can be installed in an electronic device. According to the functions implemented, the edge computing optimization system 200 of the intelligent video surveillance system may include an intelligent filtering module 201, a deviation value calculation module 202, a video enhancement module 203, an instruction generation module 204, and a report generation module 205. The module of the present invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

[0134] In this embodiment, the functions of each module / unit are as follows:

[0135] Intelligent filtering module 201, used to obtain multi-angle video streams in an intelligent video surveillance system, dynamically group the multi-angle video streams to obtain dynamic video groups, identify video frame sequences corresponding to the dynamic video groups, and intelligently filter the dynamic video groups based on the video frame sequences to obtain filtered video frames;

[0136] The deviation value calculation module 202 is used to perform real-time analysis on the filtered video frame to obtain video analysis events, wherein the video analysis events include video abnormal behavior events and specific object tracking events, generate discrete symbols and symbol strings corresponding to the video analysis events, construct a motion behavior system corresponding to the filtered video frame based on the discrete symbols and the symbol strings, and calculate the behavior deviation value corresponding to the motion behavior system and a preset standard behavior pattern;

[0137] The video enhancement module 203 is used to mark the filtered video frame based on the behavior deviation value to obtain a marked video frame, quickly render the marked video frame to obtain a rendered video, and perform video enhancement on the rendered video to obtain an enhanced video;

[0138] The instruction generation module 204 is used to transmit the enhanced video to a preset cloud center for risk assessment, obtain a risk assessment index, determine the importance level corresponding to the enhanced video based on the risk assessment index, and generate a graded feedback instruction corresponding to the multi-angle video stream based on the importance level;

[0139] The report generation module 205 is used to obtain execution feedback data after executing the hierarchical feedback instruction using the preset edge computing node, and based on the execution feedback data, perform edge optimization on the node resources in the preset edge computing node to obtain edge optimization parameters, and based on the edge optimization parameters, generate an edge optimization report corresponding to the intelligent video surveillance system.

[0140] In detail, the modules described in the edge computing optimization system 200 of the intelligent video surveillance system described in the embodiment of the present invention adopt the same technical means as the edge computing optimization method of the intelligent video surveillance system described in the accompanying drawings when used, and can produce the same technical effects, which will not be repeated here.

[0141] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. Edge computing optimization method for intelligent video surveillance system, It is characterized in that The method comprises: Acquire a multi-angle video stream in an intelligent video surveillance system, dynamically group the multi-angle video stream to obtain a dynamic video group, identify a video frame sequence corresponding to the dynamic video group, and intelligently filter the dynamic video group based on the video frame sequence to obtain a filtered video frame, wherein the intelligent filtering of the dynamic video group based on the video frame sequence to obtain the filtered video frame includes: Identify a sequence number corresponding to the video frame sequence; Based on the sequence number, adjacent frames in the dynamic video group are determined, and adjacent frame differences corresponding to the adjacent frames are calculated, wherein the calculating adjacent frame differences corresponding to the adjacent frames includes: The adjacent frame differences corresponding to the adjacent frames are calculated using the following formula: Among them, D n,n+1 represents the adjacent frame difference between the nth frame and the n+1th frame, i represents the row index of the image corresponding to the adjacent frame, N represents the number of rows of the image corresponding to the adjacent frame, j represents the column index of the image corresponding to the adjacent frame, M represents the number of columns of the image corresponding to the adjacent frame, and F n (i, j) represents the pixel value at the coordinate (i, j) of the nth frame, F n+1 (i, j) represents the pixel value at the coordinate (i, j) of the n+1th frame, ω ij represents the weight coefficient of the pixel at the coordinate (i, j), and α represents the pixel sensitivity of the image corresponding to the adjacent frame; Based on the adjacent frame differences, determining a dynamic region corresponding to the dynamic video group, and querying a filter point in the dynamic region; Based on the filtering point, intelligently filter the dynamic video group to obtain a filtered video frame; Perform real-time analysis on the filtered video frame to obtain video analysis events, wherein the video analysis events include video abnormal behavior events and specific object tracking events, generate discrete symbols and symbol strings corresponding to the video analysis events, and construct a motion behavior system corresponding to the filtered video frame based on the discrete symbols and the symbol strings, wherein the construction of the motion behavior system corresponding to the filtered video frame based on the discrete symbols and the symbol strings includes: Classify the discrete symbol and the symbol string to obtain a classification symbol; Extracting key features of symbols from the classified symbols; Initially constructing behavior patterns corresponding to the key features of the symbols, and analyzing correlations between the behavior patterns; Based on the association relationship, a motion behavior system corresponding to the filtered video frame is constructed, and a behavior deviation value corresponding to the motion behavior system and a preset standard behavior pattern is calculated; Based on the behavior deviation value, the filtered video frame is marked to obtain a marked video frame, the marked video frame is quickly rendered to obtain a rendered video, and the rendered video is enhanced to obtain an enhanced video; The enhanced video is transmitted to a preset cloud center for risk assessment to obtain a risk assessment index, and based on the risk assessment index, an importance level corresponding to the enhanced video is determined, and based on the importance level, a graded feedback instruction corresponding to the multi-angle video stream is generated, wherein the step of generating the graded feedback instruction corresponding to the multi-angle video stream based on the importance level includes: Determine a classification label corresponding to the importance classification; Formulate a feedback strategy corresponding to the multi-angle video stream based on the classification label; Querying feedback parameters in the feedback strategy; Based on the feedback parameters, generating a primary feedback instruction corresponding to the multi-angle video stream; Performing simulated execution on the primary feedback instruction to obtain simulated execution data; Based on the simulation execution data, the primary feedback instructions are hierarchically optimized to obtain hierarchical feedback instructions; After executing the hierarchical feedback instruction using the preset edge computing node, execution feedback data is obtained. Based on the execution feedback data, edge optimization is performed on the node resources in the preset edge computing node to obtain edge optimization parameters. Based on the edge optimization parameters, an edge optimization report corresponding to the intelligent video surveillance system is generated.

2. The edge computing optimization method of the intelligent video surveillance system according to claim 1, It is characterized in that The dynamically grouping the multi-angle video streams to obtain a dynamic video group includes: Performing unified format conversion on the multi-angle video streams to obtain a unified video stream; Performing scene recognition on the unified video stream to obtain a video scene; Extracting dynamic features in the video scene; Based on the dynamic features, the multi-angle video streams are dynamically grouped to obtain dynamic video groups.

3. The edge computing optimization method of the intelligent video surveillance system according to claim 1, It is characterized in that The calculating of the behavior deviation value corresponding to the sports behavior system and the preset standard behavior pattern includes: The behavior deviation value corresponding to the sports behavior system and the preset standard behavior pattern is calculated using the following formula: Wherein, Wp represents the behavior deviation value corresponding to the preset standard behavior pattern of the sports behavior system, i' represents the index of the behavior feature corresponding to the preset standard behavior pattern of the sports behavior system, N' represents the number of behavior features corresponding to the preset standard behavior pattern of the sports behavior system, w i' represents the weight of the i'th behavior feature, A i' represents the actual value of the i'th behavior feature in the motion behavior system, S i' Indicates the standard value of the i'th behavior feature in the preset standard behavior mode.

4. The edge computing optimization method of the intelligent video surveillance system according to claim 1, It is characterized in that The step of marking the filtered video frame based on the behavior deviation value to obtain a marked video frame includes: Based on a preset threshold range, the behavior deviation value is classified to obtain a classification deviation value; Identifying the abnormal behavior type corresponding to the classification deviation value; Based on the abnormal behavior type, generating a marking strategy corresponding to the filtered video frame; A video frame region corresponding to the marking strategy is determined, and the video frame region is marked to obtain a marked video frame.

5. The edge computing optimization method of the intelligent video surveillance system according to claim 1, It is characterized in that After the hierarchical feedback instruction is executed by the preset edge computing node, execution feedback data is obtained, including: Parsing the hierarchical feedback instruction to obtain parsed instruction data; Querying the execution environment corresponding to the parsed instruction data; Identifying execution tasks in the execution environment; prioritizing the execution tasks to obtain task priorities; Based on the task priority, a preset edge computing node is used to perform feedback analysis on the execution task to obtain execution feedback data.

6. Edge computing optimization system for intelligent video surveillance system, It is characterized in that The edge computing optimization method for executing the intelligent video surveillance system according to any one of claims 1 to 5 comprises: An intelligent filtering module is used to obtain a multi-angle video stream in an intelligent video surveillance system, dynamically group the multi-angle video stream to obtain a dynamic video group, identify a video frame sequence corresponding to the dynamic video group, and intelligently filter the dynamic video group based on the video frame sequence to obtain a filtered video frame, wherein the intelligent filtering of the dynamic video group based on the video frame sequence to obtain a filtered video frame includes: Identify a sequence number corresponding to the video frame sequence; Based on the sequence number, adjacent frames in the dynamic video group are determined, and adjacent frame differences corresponding to the adjacent frames are calculated, wherein the calculating adjacent frame differences corresponding to the adjacent frames includes: The adjacent frame differences corresponding to the adjacent frames are calculated using the following formula: Among them, D n,n+1 represents the adjacent frame difference between the nth frame and the n+1th frame, i represents the row index of the image corresponding to the adjacent frame, N represents the number of rows of the image corresponding to the adjacent frame, j represents the column index of the image corresponding to the adjacent frame, M represents the number of columns of the image corresponding to the adjacent frame, and F n (i, j) represents the pixel value at the coordinate (i, j) of the nth frame, F n+1 (i, j) represents the pixel value at the coordinate (i, j) of the n+1th frame, ω ij represents the weight coefficient of the pixel at the coordinate (i, j), and α represents the pixel sensitivity of the image corresponding to the adjacent frame; Based on the adjacent frame differences, determining a dynamic region corresponding to the dynamic video group, and querying a filter point in the dynamic region; Based on the filtering point, intelligently filter the dynamic video group to obtain a filtered video frame; The deviation value calculation module is used to perform real-time analysis on the filtered video frame to obtain video analysis events, wherein the video analysis events include video abnormal behavior events and specific object tracking events, generate discrete symbols and symbol strings corresponding to the video analysis events, and construct a motion behavior system corresponding to the filtered video frame based on the discrete symbols and the symbol strings, wherein the motion behavior system corresponding to the filtered video frame based on the discrete symbols and the symbol strings includes: Classify the discrete symbol and the symbol string to obtain a classification symbol; Extracting key features of symbols from the classified symbols; Initially constructing behavior patterns corresponding to the key features of the symbols, and analyzing correlations between the behavior patterns; Based on the association relationship, a motion behavior system corresponding to the filtered video frame is constructed, and a behavior deviation value corresponding to the motion behavior system and a preset standard behavior pattern is calculated; A video enhancement module, used for marking the filtered video frame based on the behavior deviation value to obtain a marked video frame, quickly rendering the marked video frame to obtain a rendered video, and performing video enhancement on the rendered video to obtain an enhanced video; An instruction generation module is used to transmit the enhanced video to a preset cloud center for risk assessment, obtain a risk assessment index, determine the importance level corresponding to the enhanced video based on the risk assessment index, and generate a graded feedback instruction corresponding to the multi-angle video stream based on the importance level, wherein the step of generating the graded feedback instruction corresponding to the multi-angle video stream based on the importance level includes: Determine a classification label corresponding to the importance classification; Formulate a feedback strategy corresponding to the multi-angle video stream based on the classification label; Querying feedback parameters in the feedback strategy; Based on the feedback parameters, generating a primary feedback instruction corresponding to the multi-angle video stream; Performing simulated execution on the primary feedback instruction to obtain simulated execution data; Based on the simulation execution data, the primary feedback instructions are hierarchically optimized to obtain hierarchical feedback instructions; A report generation module is used to obtain execution feedback data after executing the hierarchical feedback instruction using a preset edge computing node, and based on the execution feedback data, edge optimize the node resources in the preset edge computing node to obtain edge optimization parameters, and based on the edge optimization parameters, generate an edge optimization report corresponding to the intelligent video surveillance system.