Intelligent monitoring system based on intelligent industrial park

Abstract

The invention discloses an intelligent monitoring system based on an intelligent industrial park, and relates to the technical field of intelligent monitoring, and the system comprises a video data collection unit which is provided with a high-resolution high-definition camera to cover a key area of the intelligent industrial park, carries out the identity verification of a person entering the intelligent industrial park, and carries out the recognition of the person entering the intelligent industrial park. And monitoring videos of all the high-definition cameras are obtained, and the obtained monitoring videos are processed. The invention provides an intelligent monitoring system based on an intelligent industrial park, and the system carries out the identity verification of a person entering the intelligent industrial park through an arranged video data collection unit, obtains the monitoring video of each high-definition camera, carries out the processing of the obtained monitoring video, and carries out the processing of the obtained monitoring video. According to the method, a monitoring video is set to obtain comprehensive information expression of employee behaviors in the monitoring video, an abnormal behavior detection network is designed through a set abnormal behavior recognition unit, a training data set is constructed, and the abnormal behavior detection network is trained through the training data set.

Description

Technical Field

[0001] The present invention relates to the field of intelligent monitoring technology, and in particular to an intelligent monitoring system based on a smart industrial park. Background Art

[0002] Smart industrial parks are a new type of industrial park model that uses advanced information technology, the Internet of Things, big data, artificial intelligence and other modern scientific and technological means to carry out intelligent transformation and upgrading of traditional industrial parks to improve management efficiency, production efficiency and environmental sustainability. The Chinese patent with application number 202310615985.8 discloses "an industrial park carbon emission intelligent monitoring system and method, which relates to the field of carbon emission monitoring technology, including: obtaining relevant data of the industrial park and marking the relevant data of the industrial park; then using the marked relevant data of the industrial park to calculate the production coefficient, and monitoring and counting the daily production power of each industrial enterprise in the industrial park to obtain the enterprise electricity consumption data, and then using the calculated production coefficient and enterprise electricity consumption data to calculate the carbon emission calculation amount, and then collecting the total carbon emission, using the obtained carbon emission calculation amount and the total carbon emission for comprehensive judgment, using the intersection method to obtain different situations, and analyzing whether there is a problem with carbon emission according to different situations. If there is a problem, the problem is located in production or enterprise electricity consumption, and an alarm is issued according to the problem and a corresponding solution is provided. ";

[0003] This prior art only solves the problem that when implementing existing carbon emission monitoring solutions for industrial parks, most of them only monitor the emissions of various industrial enterprises based on monitoring equipment, which means that carbon emissions in industrial parks can only be monitored in a single way, and cannot be monitored and analyzed in a diversified way from different aspects, resulting in the overall poor effect of carbon emission monitoring in industrial parks. It does not take into account that when monitoring a smart industrial park, the employee's work area should be determined first, so as to facilitate the subsequent detection of employee area intrusion behavior and track the employee's movement trajectory, so as to facilitate timely handling after notifying security personnel, and based on surveillance videos, employees who have committed assaults can be detected to protect their physical safety. Summary of the invention

[0004] The purpose of the present invention is to provide an intelligent monitoring system based on a smart industrial park to solve the problems raised in the above background technology.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: an intelligent monitoring system based on a smart industrial park, comprising a video data collection unit, wherein the video data collection unit deploys high-resolution HD cameras to cover key areas of the smart industrial park, and performs identity authentication on personnel entering the smart industrial park, and obtains monitoring videos of each HD camera, and processes the obtained monitoring videos;

[0006] An abnormal behavior recognition unit, wherein the abnormal behavior recognition unit is designed to detect abnormal behavior of personnel in combination with human skeleton information, and a training data set is constructed to train the abnormal behavior detection network, and an accuracy verification operation is performed on the trained abnormal behavior detection network, and then the abnormal behavior information existing in the processed surveillance video is outputted by the abnormal behavior detection network;

[0007] An intelligent monitoring and analysis unit, which is designed to detect abnormal personnel intruding into the area, designs a regional intrusion detection network, obtains the detection results of the regional intrusion detection network, and then tracks the corresponding personnel, thereby avoiding safety accidents caused by personnel approaching dangerous areas;

[0008] The analysis result display unit displays real-time monitoring information and detection result information in the form of a visual dashboard through the monitoring platform, helping managers to fully understand the personnel dynamics and potential risks in the park, and notifying security personnel through multiple channels when abnormal alarms occur to ensure timely response.

[0009] Preferably, the video data collection unit includes a monitoring equipment deployment module and a personnel identity authentication module, the monitoring equipment deployment module deploys high-resolution HD cameras in key areas of the smart industrial park, and associates each high-resolution HD camera with a corresponding monitoring area, and marks the dangerous areas therein in red, the personnel identity authentication module and the monitoring equipment deployment module are electrically connected, the personnel identity authentication module is provided with an employee information database, and the employee information database stores basic identity information, facial information and work area information of each employee, and when employees resign and new employees join, the employee information database is updated in time, and when employees enter the smart industrial park, the facial information of the employees is obtained and compared through the facial recognition set in the intelligent access control system, thereby realizing the identity authentication of the employees, and further based on whether the facial information of all employees has been called for comparison and the comparison results, the full attendance status of the smart industrial park on that day is determined.

[0010] Preferably, the video data collection unit also includes a monitoring video acquisition module and a monitoring video processing module, the monitoring video acquisition module is electrically connected to the personnel identity authentication module, the monitoring video acquisition module acquires monitoring videos of various monitoring areas through a high-definition camera, the monitoring video processing module is electrically connected to the monitoring video acquisition module, and the monitoring video processing module processes the acquired monitoring videos to obtain a more comprehensive information expression of employee behavior in the monitoring video. The specific operation is to achieve aggregation of input video spatial position information and motion information by constructing attention at three levels: space, channel, and motion. The specific operation of constructing spatial attention is to compress channel information through channel pooling operation, thereby focusing on key areas in the space, acquiring local information in the monitoring video, and achieving spatial The construction of attention, where the specific operation of constructing channel attention is to extract features from the input surveillance video, obtain feature maps, and perform channel segmentation on it, so as to obtain feature blocks with channel attention, and then use multiple bottleneck residual blocks to fuse the feature blocks containing feature information of different scales after processing, so as to obtain a wider global receptive field and a more comprehensive feature representation, and realize global modeling of channel information. The specific operation of constructing motion attention is to extract features from the input surveillance video, obtain feature matrices, and split them into time dimensions. One part is used to learn the difference in object motion in the previous and next time periods, and the other part is used to highlight the position of the object in the feature map. Then the position information is combined with the motion information, and the encoding-pooling-decoding structure is used to realize the construction of motion attention.

[0011] Preferably, the abnormal behavior identification unit includes an abnormal behavior detection network design module and a training data set construction module. The abnormal behavior detection network design module is designed for the purpose of detecting abnormal behaviors of beating people. The abnormal behavior detection network is designed based on the OpenPose human posture estimation model in combination with human skeleton information. The skeleton data is extracted through the original OpenPose human posture estimation model, and the skeleton data features are submitted to the LighterFace lightweight detection network for identification. The specific operation is to find all hands and feet in each frame of the picture through the Bottom-Up method in the OpenPose algorithm, encode the position and direction of the limbs in each frame of the image through the partial region affinity branch in the OpenPose human posture estimation model, and then mark the confidence of each body part. Combined with these two branches, the position and connection of the key points are extracted. When the key points of the human body in each frame of the image are detected, the line integral on the line segment between the key points of the human body is calculated. The specific calculation formula is as follows:

[0012]

[0013] Where, L C Indicates the length of the body part, dj1 Indicates body part j1 , d j2 Indicates body part j2 Based on the calculation results, the correlation between the key points of the body parts is inferred to determine the whole body posture of the person, and then the abnormal behavior detection of the beating of the person is realized. The training data set construction module crawls the video samples of the beating of the person through the web crawler technology, and processes the crawled video samples of the beating of the person, and constructs the training data set based on the processed video samples of the beating of the person.

[0014] Preferably, the abnormal behavior identification unit also includes an abnormal behavior network training module and an abnormal behavior network output module, the abnormal behavior network training module is electrically connected to the abnormal behavior detection network design module and the training data set construction module respectively, the abnormal behavior network training module trains the abnormal behavior detection network through the training data set, and performs accuracy verification operation on the trained abnormal behavior detection network, the abnormal behavior network output module is electrically connected to the abnormal behavior network training module, the abnormal behavior network output module outputs the abnormal behavior information of beating in the processed surveillance video through the abnormal behavior detection network, and obtains the identity information and location information of the corresponding employees.

[0015] Preferably, the intelligent monitoring and analysis unit includes a regional intrusion detection network design module, which aims to detect abnormal personnel in the intrusion area and designs a regional intrusion detection network based on the YOLOv5 target detection network. Specifically, an AGPA attention model is added between the last layer of the original YOLOv5 target detection backbone network and the previous level of the SPP module. The added AGPA attention model includes a global feature extraction module and an adaptive learning module, wherein the global feature extraction module uses one-dimensional convolution to extract two global pooling features on the channel dimension, so that the amount of network parameters is greatly reduced. In the adaptive learning module, learnable parameters are used to adaptively adjust the proportion of the two global pooling features and fuse them to obtain more comprehensive and discriminative channel attention features. The added AGPA attention model guides the regional intrusion detection network to focus more effectively on the key features of the personnel target, enhances the recognition of detail features by the regional intrusion detection network, and thus improves the accuracy of the regional intrusion detection network.

[0016] Preferably, the intelligent monitoring and analysis unit also includes a detection result acquisition module and an intruder tracking module. The detection result acquisition module is electrically connected to the regional intrusion detection network design module. The detection result acquisition module acquires the personnel detection results of the regional intrusion detection network on the monitoring videos of each area, obtains the personnel information in the monitoring videos of each area, and compares the area information corresponding to the monitoring video with the work area information corresponding to the personnel. Based on the comparison result, the employee information with abnormal regional intrusion behavior is determined. The abnormal regional intrusion behavior is specifically defined as an employee appearing in a work area that does not belong to him during working hours, and the employee information of invading a dangerous area is marked in red. The intruder tracking module is electrically connected to the detection result acquisition module. The intruder tracking module tracks the employees with abnormal regional intrusion behavior, specifically by associating the location information of the target employee in each monitoring video area, determining the movement trajectory of the target employee, and tracking the target employee.

[0017] Preferably, the analysis result display unit includes an information visualization display module and an abnormal alarm response module, the information visualization display module integrates the monitoring video, the identity information and location information of the employees with abnormal beating behavior, the information of the employees with abnormal regional invasion behavior and the movement trajectory information of the employees through the Internet of Things, and displays the above information in the form of a visual dashboard through the monitoring platform, the abnormal alarm response module and the information visualization display module are electrically connected, the abnormal alarm response module notifies the security personnel of the identity information and location information of the employees with abnormal beating behavior and the information of the employees with abnormal regional invasion behavior through multiple channels to ensure timely response, and based on the personnel detection results of the monitoring videos of each area by the regional intrusion detection network, when there are staff members belonging to the area who have not appeared in the monitoring area for a long time, the early warning mechanism will be automatically triggered to notify the management personnel in time.

[0018] Compared with the prior art, the beneficial effects of the present invention at least include: the present invention proposes an intelligent monitoring system based on a smart industrial park, which authenticates the identity of persons entering the smart industrial park through a set video data collection unit, obtains monitoring videos of various high-definition cameras, and processes the obtained monitoring videos to obtain a more comprehensive information expression of employee behavior in the monitoring videos, designs an abnormal behavior detection network through a set abnormal behavior recognition unit, constructs a training data set, trains the abnormal behavior detection network through the training data set, outputs the information of abnormal beating behavior in the processed monitoring video through the abnormal behavior detection network, and obtains identity information and location information of the corresponding employees, designs a regional intrusion detection network through a set intelligent monitoring analysis unit, determines employee information with regional intrusion abnormal behavior based on the detection results, determines the movement trajectory of the target employee by associating the location information of the target employee in each monitoring video area, and realizes tracking of the target employee, displays the monitoring information in the form of a visual dashboard through a set analysis result display unit, and notifies security personnel of the information of employees with abnormal behavior through multiple channels to ensure timely response. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the overall process of an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of the process of a video data collection unit according to an embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of the abnormal behavior identification unit flow in an embodiment of the present invention;

[0022] Figure 4 This is a schematic diagram of the process of the intelligent monitoring and analysis unit according to an embodiment of the present invention;

[0023] Figure 5 This is a schematic diagram of the analysis result display unit flow in an embodiment of the present invention.

[0024] In the figure: 100, video data collection unit; 101, monitoring equipment deployment module; 102, personnel identity authentication module; 103, monitoring video acquisition module; 104, monitoring video processing module; 200, abnormal behavior recognition unit; 201, abnormal behavior detection network design module; 202, training data set construction module; 203, abnormal behavior network training module; 204, abnormal behavior network output module; 300, intelligent monitoring and analysis unit; 301, regional intrusion detection network design module; 302, detection result acquisition module; 303, intruder tracking module; 400, analysis result display unit; 401, information visualization display module; 402, abnormal alarm response module. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] See also Figure 1-Figure 5 , the present invention provides a technical solution: an intelligent monitoring system based on a smart industrial park, comprising a video data collection unit 100, the video data collection unit 100 deploys high-resolution HD cameras to cover key areas of the smart industrial park, and authenticates personnel entering the smart industrial park, obtains monitoring videos from each HD camera, and processes the acquired monitoring videos;

[0027] The abnormal behavior recognition unit 200 is designed for the purpose of detecting abnormal behavior of personnel, combines human skeleton information to design an abnormal behavior detection network, and constructs a training data set to train the abnormal behavior detection network, and performs accuracy verification operations on the trained abnormal behavior detection network, and then outputs abnormal behavior information in the processed surveillance video through the abnormal behavior detection network;

[0028] The intelligent monitoring and analysis unit 300 is designed to detect abnormal persons invading the area, and obtains the detection results of the regional intrusion detection network, and then tracks the corresponding persons, so as to avoid safety accidents caused by persons approaching the dangerous area;

[0029] The analysis result display unit 400 displays real-time monitoring information and detection result information in the form of a visual dashboard through the monitoring platform, helping managers to fully understand the personnel dynamics and potential risks in the park, and notify security personnel through multiple channels when abnormal alarms occur to ensure timely response.

[0030] The video data collection unit 100 includes a monitoring equipment deployment module 101 and a personnel identity authentication module 102. The monitoring equipment deployment module 101 deploys high-resolution HD cameras in key areas of the smart industrial park, and associates each high-resolution HD camera with a corresponding monitoring area, and marks the dangerous area in red. The personnel identity authentication module 102 is electrically connected to the monitoring equipment deployment module 101. The personnel identity authentication module 102 is provided with an employee information database, and the employee information database stores basic identity information, facial information and work area information of each employee. When an employee resigns or a new employee joins, the employee information database is updated in a timely manner. When an employee enters the smart industrial park, the facial information of the employee is obtained and compared through the facial recognition set in the intelligent access control system, thereby realizing the identity authentication of the employee, and further based on whether the facial information of all employees is called for comparison and the comparison results, the full attendance of the smart industrial park on that day is determined;

[0031] The video data collection unit 100 also includes a monitoring video acquisition module 103 and a monitoring video processing module 104. The monitoring video acquisition module 103 is electrically connected to the personnel identity authentication module 102. The monitoring video acquisition module 103 acquires monitoring videos of various monitoring areas through a high-definition camera. The monitoring video processing module 104 is electrically connected to the monitoring video acquisition module 103. The monitoring video processing module 104 processes the acquired monitoring videos to obtain a more comprehensive information expression of employee behavior in the monitoring videos. The specific operation is to achieve the aggregation of input video spatial position information and motion information by constructing attention at three levels: space, channel, and motion. The specific operation of constructing spatial attention is to compress channel information through channel pooling operation, thereby focusing on key areas in the space and acquiring local information in the monitoring video. Information is obtained to realize the construction of spatial attention. The specific operation of constructing channel attention is to extract features from the input surveillance video, obtain feature maps, and perform channel segmentation on them, so as to obtain feature blocks with channel attention. Then, multiple bottleneck residual blocks are used to fuse the feature blocks containing feature information of different scales after processing, so as to obtain a wider global receptive field and a more comprehensive feature representation, so as to realize the global modeling of channel information. The specific operation of constructing motion attention is to extract features from the input surveillance video, obtain feature matrices, and split them in the time dimension. One part is used to learn the difference of object motion in the previous and next time periods, and the other part is used to highlight the position of the object in the feature map. Then, the position information is combined with the motion information, and the encoding-pooling-decoding structure is used to realize the construction of motion attention.

[0032] The abnormal behavior identification unit 200 includes an abnormal behavior detection network design module 201 and a training data set construction module 202. The abnormal behavior detection network design module 201 is designed for the purpose of detecting abnormal behaviors of beating people. The abnormal behavior detection network is designed based on the OpenPose human posture estimation model in combination with human skeleton information. The skeleton data is extracted through the original OpenPose human posture estimation model, and the skeleton data features are submitted to the LighterFace lightweight detection network for identification. The specific operation is to find all hands and feet in each frame of the picture through the Bottom-Up method in the OpenPose algorithm, encode the position and direction of the limbs in each frame of the image through the partial region affinity branch in the OpenPose human posture estimation model, and then mark the confidence of each body part. Combined with these two branches, the position and connection of the key points are extracted. When the key points of the human body in each frame of the image are detected, the line integral on the line segment between the key points of the human body is calculated. The specific calculation formula is as follows:

[0033]

[0034] Where, L C Indicates the length of the body part, d j1 Indicates body part j1 , d j2 Indicates body part j2 Based on the calculation results, the correlation between the key points of the body parts is inferred, so as to determine the whole body posture of the person, and then realize the detection of abnormal beating behavior of the person. The training data set construction module 202 crawls the video samples of the beating of the person through the web crawler technology, and processes the crawled video samples of the beating of the person, and constructs the training data set based on the processed video samples of the beating of the person;

[0035] The abnormal behavior identification unit 200 further includes an abnormal behavior network training module 203 and an abnormal behavior network output module 204. The abnormal behavior network training module 203 is electrically connected to the abnormal behavior detection network design module 201 and the training data set construction module 202 respectively. The abnormal behavior network training module 203 trains the abnormal behavior detection network through the training data set and performs accuracy verification operation on the trained abnormal behavior detection network. The abnormal behavior network output module 204 is electrically connected to the abnormal behavior network training module 203. The abnormal behavior network output module 204 outputs the abnormal behavior information of beating in the processed surveillance video through the abnormal behavior detection network, and obtains the identity information and location information of the corresponding employee.

[0036] The intelligent monitoring and analysis unit 300 includes a regional intrusion detection network design module 301. The regional intrusion detection network design module 301 is designed for the purpose of detecting abnormal personnel in the intrusion area. The regional intrusion detection network is designed based on the YOLOv5 target detection network. Specifically, an AGPA attention model is added between the last layer of the original YOLOv5 target detection backbone network and the previous level of the SPP module. The added AGPA attention model includes a global feature extraction module and an adaptive learning module. The global feature extraction module uses a one-dimensional convolution to extract two global pooling features on the channel dimension, so that the network parameters are greatly reduced. In the adaptive learning module, the proportion of the two global pooling features is adaptively adjusted and fused using learnable parameters to obtain a more comprehensive and discriminative channel attention feature. The added AGPA attention model guides the regional intrusion detection network to focus more effectively on the key features of the personnel target, enhances the recognition of the detailed features by the regional intrusion detection network, and thus improves the accuracy of the regional intrusion detection network.

[0037] The intelligent monitoring and analysis unit 300 also includes a detection result acquisition module 302 and an intruder tracking module 303. The detection result acquisition module 302 is electrically connected to the regional intrusion detection network design module 301. The detection result acquisition module 302 acquires the personnel detection results of the regional intrusion detection network on the monitoring videos of each region, obtains the personnel information in the monitoring videos of each region, and compares the regional information corresponding to the monitoring video with the corresponding work area information of the personnel. Based on the comparison result, the employee information with abnormal regional intrusion behavior is determined. The abnormal regional intrusion behavior is specifically defined as the employee appearing in the work area that does not belong to him during working hours, and the employee information of the employee invading the dangerous area is marked in red. The intruder tracking module 303 is electrically connected to the detection result acquisition module 302. The intruder tracking module 303 tracks the employees with abnormal regional intrusion behavior, specifically by associating the location information of the target employee in each monitoring video area, determining the movement trajectory of the target employee, and realizing the tracking of the target employee;

[0038] The analysis result display unit 400 includes an information visualization display module 401 and an abnormal alarm response module 402. The information visualization display module 401 integrates the monitoring video, the identity information and location information of the employees with abnormal beating behavior, the information of the employees with abnormal regional intrusion behavior, and the movement trajectory information of the employees through the Internet of Things, and displays the above information in the form of a visual dashboard through the monitoring platform. The abnormal alarm response module 402 and the information visualization display module 401 are electrically connected. The abnormal alarm response module 402 notifies the security personnel of the identity information and location information of the employees with abnormal beating behavior and the information of the employees with abnormal regional intrusion behavior through multiple channels to ensure timely response. Based on the personnel detection results of the monitoring videos of each area by the regional intrusion detection network, when there are staff members belonging to the area who have not appeared in the monitoring area for a long time, the early warning mechanism will be automatically triggered to notify the management personnel in time.

[0039] Working principle: High-resolution HD cameras are deployed in key areas of the smart industrial park through the monitoring equipment deployment module 101, and each high-resolution HD camera is associated with the corresponding monitoring area. The identity of employees is authenticated through the personnel identity verification module 102, and the monitoring video of each monitoring area is acquired through the monitoring video acquisition module 103. The acquired monitoring video is processed by the monitoring video processing module 104 to obtain a more comprehensive information expression of employee behavior in the monitoring video. The abnormal behavior detection network is designed through the abnormal behavior detection network design module 201, and the training data set is constructed through the training data set construction module 202. The abnormal behavior detection network is trained through the abnormal behavior network training module 203, and the accuracy verification operation is performed on the trained abnormal behavior detection network. The behavior network output module 204 outputs the information of abnormal beating behavior in the processed surveillance video, and obtains the identity information and location information of the corresponding employees. The regional intrusion detection network is designed through the regional intrusion detection network design module 301. The personnel detection results of the regional intrusion detection network for the surveillance videos of each area are obtained through the detection result acquisition module 302, and the personnel information in the surveillance videos of each area is obtained. The movement trajectory of the target employee is determined through the intruder tracking module 303 to achieve tracking of the target employee. The monitoring information is displayed in the form of a visual dashboard through the information visualization display module 401. The identity information and location information of employees with abnormal beating behavior and the information of employees with abnormal regional intrusion behavior are notified to security personnel through multiple channels through the abnormal alarm response module 402 to ensure timely response.

[0040] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0041] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent monitoring system based on a smart industrial park, characterized by: The device comprises a video data collection unit (100), wherein the video data collection unit (100) deploys high-resolution HD cameras to cover key areas of a smart industrial park, performs identity verification on persons entering the smart industrial park, obtains monitoring videos from each HD camera, and processes the obtained monitoring videos; An abnormal behavior recognition unit (200) is designed for the purpose of detecting abnormal behavior of a person, combines human skeleton information to design an abnormal behavior detection network, constructs a training data set to train the abnormal behavior detection network, performs accuracy verification operations on the trained abnormal behavior detection network, and then outputs abnormal behavior information present in the processed surveillance video through the abnormal behavior detection network; An intelligent monitoring and analysis unit (300), wherein the intelligent monitoring and analysis unit (300) is designed to detect abnormal persons invading an area, and to obtain detection results of the regional intrusion detection network, thereby tracking the corresponding persons, thereby avoiding safety accidents caused by persons approaching dangerous areas; The analysis result display unit (400) displays real-time monitoring information and detection result information in the form of a visual dashboard through a monitoring platform, helping management personnel to fully understand the personnel dynamics and potential risks in the park, and notifying security personnel through multiple channels when abnormal alarms occur to ensure timely response.

2. According to claim 1, the intelligent monitoring system based on the smart industrial park is characterized in that: The video data collection unit (100) comprises a monitoring device deployment module (101) and a personnel identity authentication module (102). The monitoring device deployment module (101) deploys high-resolution HD cameras in key areas of the smart industrial park, and associates each high-resolution HD camera with a corresponding monitoring area, and marks the dangerous area therein in red. The personnel identity authentication module (102) is electrically connected to the monitoring device deployment module (101). The personnel identity authentication module (102) is provided with an employee information database, and the employee information database stores basic identity information, facial information and work area information of each employee. When an employee resigns or a new employee joins, the employee information database is updated in a timely manner. When an employee enters the smart industrial park, the facial information of the employee is obtained and compared through the facial recognition set in the intelligent access control system, thereby realizing the identity authentication of the employee. The full attendance of the smart industrial park on that day is further determined based on whether the facial information of all employees has been called for comparison and the comparison result.

3. According to claim 2, the intelligent monitoring system based on the smart industrial park is characterized in that: The video data collection unit (100) further comprises a monitoring video acquisition module (103) and a monitoring video processing module (104); the monitoring video acquisition module (103) is electrically connected to the personnel identity verification module (102); the monitoring video acquisition module (103) acquires monitoring videos of various monitoring areas through a high-definition camera; the monitoring video processing module (104) is electrically connected to the monitoring video acquisition module (103); and the monitoring video processing module (104) processes the acquired monitoring videos to obtain a more comprehensive information expression of employee behavior in the monitoring videos.

4. According to claim 1, the intelligent monitoring system based on the smart industrial park is characterized in that: The abnormal behavior recognition unit (200) comprises an abnormal behavior detection network design module (201) and a training data set construction module (202). The abnormal behavior detection network design module (201) is designed for the purpose of detecting abnormal behaviors of beating people, and combines human skeleton information with an OpenPose human posture estimation model to design an abnormal behavior detection network. Skeleton data is extracted through the original OpenPose human posture estimation model, and the skeleton data features are submitted to the LighterFace lightweight detection network for recognition. The specific operation is to find all hands and feet in each frame of the picture through the Bottom-Up method in the OpenPose algorithm, and to obtain the abnormal behavior detection network through the OpenPose algorithm. The partial region affinity branch in the ose human posture estimation model encodes the position and direction of the limbs in each frame image, and then marks the confidence of each body part. The two branches are combined to extract features of the position and connection of key points. When the human key points in each frame image are detected, the line integral on the line segment between the human key points is calculated, and the correlation between the key points of the body parts is inferred based on the calculation results, so as to determine the whole body posture of the person, and then realize the detection of abnormal beating behavior of people. The training data set construction module (202) crawls the video samples of people beating by using the web crawler technology, processes the crawled video samples of people beating, and constructs the training data set based on the processed video samples of people beating.

5. According to claim 4, the intelligent monitoring system based on the smart industrial park is characterized in that: The abnormal behavior identification unit (200) further comprises an abnormal behavior network training module (203) and an abnormal behavior network output module (204); the abnormal behavior network training module (203) is electrically connected to the abnormal behavior detection network design module (201) and the training data set construction module (202), respectively; the abnormal behavior network training module (203) trains the abnormal behavior detection network through the training data set, and performs an accuracy verification operation on the trained abnormal behavior detection network; the abnormal behavior network output module (204) is electrically connected to the abnormal behavior network training module (203); the abnormal behavior network output module (204) outputs the abnormal beating abnormal behavior information present in the processed surveillance video through the abnormal behavior detection network, and obtains the identity information and location information of the corresponding employee.

6. The intelligent monitoring system based on a smart industrial park according to claim 1 is characterized in that: The intelligent monitoring and analysis unit (300) comprises a regional intrusion detection network design module (301). The regional intrusion detection network design module (301) is designed for the purpose of detecting abnormal personnel in the intrusion area based on the YOLOv5 target detection network. Specifically, an AGPA attention model is added between the last layer of the original YOLOv5 target detection backbone network and the previous level of the SPP module. The added AGPA attention model comprises a global feature extraction module and an adaptive learning module. The global feature extraction module uses one-dimensional convolution to extract two global pooling features on the channel dimension, so that the network parameter quantity is greatly reduced. In the adaptive learning module, learnable parameters are used to adaptively adjust the proportion of the two global pooling features and fuse them to obtain more comprehensive and more discriminative channel attention features. The added AGPA attention model guides the regional intrusion detection network to more effectively focus on the key features of the personnel target, enhances the recognition of the detailed features by the regional intrusion detection network, and thus improves the accuracy of the regional intrusion detection network.

7. The intelligent monitoring system based on a smart industrial park according to claim 6 is characterized in that: The intelligent monitoring and analysis unit (300) further comprises a detection result acquisition module (302) and an intruder tracking module (303). The detection result acquisition module (302) is electrically connected to the regional intrusion detection network design module (301). The detection result acquisition module (302) acquires the personnel detection results of the regional intrusion detection network on the monitoring videos of each region, obtains the personnel information in the monitoring videos of each region, and compares the regional information corresponding to the monitoring videos with the work area information corresponding to the personnel, and determines the information of employees with abnormal regional intrusion behavior based on the comparison result. The abnormal regional intrusion behavior is specifically defined as an employee appearing in a work area that does not belong to him during working hours, and the information of employees who invade the dangerous area is marked in red. The intruder tracking module (303) is electrically connected to the detection result acquisition module (302). The intruder tracking module (303) tracks the employees with abnormal regional intrusion behavior, specifically by associating the location information of the target employee appearing in each monitoring video area, determining the movement trajectory of the target employee, and realizing the tracking of the target employee.

8. The intelligent monitoring system based on a smart industrial park according to claim 1 is characterized in that: The analysis result display unit (400) comprises an information visualization display module (401) and an abnormal alarm response module (402). The information visualization display module (401) integrates the monitoring video, the identity information and location information of the employees with abnormal beating behavior, the information of the employees with abnormal regional intrusion behavior, and the movement trajectory information of the employees through the Internet of Things, and displays the above information in the form of a visual dashboard through the monitoring platform. The abnormal alarm response module (402) and the information visualization display module (401) are electrically connected. The abnormal alarm response module (402) notifies the security personnel through multiple channels of the identity information and location information of the employees with abnormal beating behavior and the information of the employees with abnormal regional intrusion behavior to ensure timely response. Based on the personnel detection results of the monitoring videos of each area by the regional intrusion detection network, when there is a staff member belonging to the area who has not appeared in the monitoring area for a long time, the early warning mechanism will be automatically triggered to notify the management personnel in time.

Images