Intelligent monitoring system for operation construction of chemical enterprise

Abstract

The invention discloses an intelligent monitoring system for operation construction of a chemical enterprise, and relates to the technical field of safety supervision of special operation, and the system comprises an equipment layer which comprises special operation intelligent guardian equipment and special operation personnel positioning equipment, and is used for providing data support for an algorithm layer and a platform layer; the algorithm layer is used for constructing a worker behavior intelligent analysis model and a worker real-time positioning and supervision model; and the platform layer comprises a video management and AI intelligent analysis system and a special operation permission and operation process management system. According to the invention, intelligent guardian equipment is taken as a core, technologies of artificial intelligence, personnel positioning and the like are fused, an intelligent supervision technology for special operation of a chemical enterprise is explored and researched, and meanwhile, a software platform is combined, so that whole-process licensing management of operation, video supervision of field operation and continuous sampling analysis of toxic and harmful gas are realized; and non-standard behaviors and unsafe scenes in the whole operation process are accurately identified and pre-warned by using technologies such as visual analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of safety supervision for special operations, and more specifically, to an intelligent monitoring system for construction operations in chemical enterprises. Background Art

[0002] During the production process of chemical enterprises, a large number of toxic, harmful, flammable, explosive and other dangerous chemical substances are involved, which are extremely likely to cause safety accidents such as fires, explosions, and poisonings, seriously threatening the lives of operating personnel and the property safety of enterprises. As a weak link in the safety management of chemical enterprises, special operations are extremely prone to various safety accidents such as mechanical injuries, lifting injuries, burns, object strikes, falls from heights, and fire explosions due to their complex and changeable construction environments and the uneven professional capabilities and practical experiences of operating personnel.

[0003] Therefore, it is particularly important to implement precise and effective safety supervision measures during the operation process. At present, the safety supervision of special operations in chemical enterprises mainly relies on traditional means such as safety education and training and on-site manual monitoring, but there are many significant limitations: for example, the efficiency of manual monitoring is low, and it is difficult to achieve real-time dynamic monitoring of the operating environment; the existing supervision technologies are difficult to adapt to the complex and changeable chemical environment and lack intelligence and automation capabilities; the collection and analysis of safety data are insufficient, making it difficult to provide strong data support for management decisions. In addition, there are still common problems in the management of special operations in chemical enterprises at present, such as non-compliant approval processes, failure of guardians to perform their duties, and failure to implement safety measures in place. Although some research has made certain progress in addressing these problems, most research still focuses on visual monitoring and is difficult to timely identify and warn of on-site problems through technologies such as artificial intelligence, resulting in obvious deficiencies in the comprehensive supervision ability of the operation site.

[0004] Therefore, how to propose an intelligent monitoring system for construction operations in chemical enterprises, build a full-process supervision system for special operations in chemical enterprises, build a special operation intelligent guardian system integrating continuous gas sampling, real-time video monitoring, and intelligent analysis of on-site personnel behavior, integrate and analyze real-time monitoring data, and improve the safety supervision efficiency and management level of special operations is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides an intelligent monitoring system for construction operations in chemical enterprises, builds a full-process supervision system for special operations in chemical enterprises, builds a special operation intelligent guardian system integrating continuous gas sampling, real-time video monitoring, and intelligent analysis of on-site personnel behavior, integrates and analyzes real-time monitoring data, and improves the safety supervision efficiency and management level of special operations. To achieve the above object, the present invention adopts the following technical solutions:

[0006] An intelligent monitoring system for chemical enterprise operation construction, comprising:

[0007] Device layer: including special operation intelligent guardian devices and special operation personnel positioning devices, used to provide data support for the algorithm layer and the platform layer;

[0008] Algorithm layer: used to build an intelligent analysis model for the behavior of operation personnel and a real-time positioning and supervision model for operation personnel;

[0009] Platform layer: including a video management and AI intelligent analysis system and a special operation permission and operation process management system. The video management and AI intelligent analysis system is used to store and manage the video data collected during the operation process, and at the same time call the models of the algorithm layer for data analysis. The special operation permission and operation process management system is used to provide the full-process management functions of special operation application, process approval, execution monitoring, and post-event analysis.

[0010] Optionally, the special operation intelligent guardian device includes: an explosion-proof enclosure, a main control board, a pump suction system, a video acquisition system, a battery charging management module, a wireless communication module, an audible and visual alarm module, a digital display module, and a mobile terminal Bluetooth configuration module; the main control board, the pump suction system, the battery charging management module, the wireless communication module, the audible and visual alarm module, the digital display module, and the mobile terminal Bluetooth configuration module are placed inside the explosion-proof enclosure, and the main control board is respectively signal-connected to the pump suction system, the video acquisition system, the battery charging management module, the wireless communication module, the audible and visual alarm module, the digital display module, and the mobile terminal Bluetooth configuration module.

[0011] Optionally, the pump suction system includes an air inlet, an air outlet, an air pump, and a gas sensor. The air inlet, the gas sensor, the air pump, and the air outlet are connected in sequence, and the gas sensor is signal-connected to the main control board.

[0012] Optionally, the video management and AI intelligent analysis system includes: a device management module, an algorithm configuration module, a real-time monitoring module, a historical video module, and an intelligent analysis module;

[0013] The device management module: includes hierarchical and classification management of intelligent guardian devices, maintains the basic information and bound power-on and power-off information of the devices, and provides an overview of the device status;

[0014] The algorithm configuration module: used to configure and manage the intelligent analysis algorithm for the behavior of operation personnel;

[0015] The real-time monitoring module: provides a split-screen display function, used to simultaneously provide users with real-time monitoring images transmitted by multiple intelligent guardian devices and fixed cameras;

[0016] The historical video module: used to store and manage the video data collected by the intelligent guardian device and fixed cameras during the operation process, and also supports video playback;

[0017] The intelligent analysis module: used to perform real-time analysis on the behaviors of personnel at the operation site based on the personnel behavior recognition algorithm, generate warning information when abnormal situations are detected, and feedback it to the intelligent guardian device for operations and the special operation permit and operation process management system.

[0018] Optionally, the special operation permit and operation process management system includes: a permit management module, an operation implementation module, an operation monitoring module, an operation alarm module, and a statistical analysis module;

[0019] The permit management module: used to provide functions for operation application, safety measure confirmation, sampling analysis, and approval of operation permits. After an operation application, the system automatically pushes tasks to relevant personnel for measure confirmation, sampling analysis, and approval according to the process. The relevant personnel perform corresponding operations through the special operation mobile terminal and upload the results. At the same time, the system will automatically associate operation-related information, generate an electronic operation ticket according to the standard operation ticket template, and support preview and download of the operation ticket;

[0020] The operation implementation module: used to provide functions for operation start, pause, resume, end, and safety disclosure operation implementation management. At the same time, when the operation starts, it supports scanning the code to bind the intelligent guardian device and supports viewing real-time alarm information during the operation process;

[0021] The operation monitoring module: supports managers to remotely view real-time information at the operation site;

[0022] The operation alarm module: used to integrate various warning analysis algorithms for alarm analysis, notify relevant personnel of the alarm information, and support tracking the progress of alarm handling;

[0023] The statistical analysis module: used to provide functions for operation and alarm statistical analysis, and summarize and analyze the on-site operation tickets and alarm situations according to dimensions such as time, operation ticket type, applying unit, and alarm type.

[0024] Optionally, the real-time positioning and supervision model for operating personnel includes: inputting the real-time monitoring image into the target detection algorithm to detect the human body in the image and mark the position information; judging whether an employee is absent from work based on the human body and the position information. If no human body target is detected in a certain employee's area for more than a preset time threshold, then it is determined that the employee has left the post; if the number of human bodies exceeds the preset number threshold when the human body is marked in a certain area through the target detection algorithm, then it is determined that the area is overstaffed.

[0025] Optionally, it further includes: when a human target is detected in the area, identifying the human body and markers in the area; determining the correlation between the human body and the markers; judging whether the human body in the target area is the target object according to the correlation; and determining the off-duty status of the employees in the area according to the judgment result.

[0026] Optionally, the determining the correlation between the human body and the markers includes: determining the human body area occupied by the human body, and determining the marker area occupied by the markers; determining the coincidence situation between the human body area and the marker area, where the coincidence situation is the ratio of the area of the overlapping area between the human body area and the marker area to the area of the marker area; and determining the correlation between the human body and the markers based on the coincidence situation.

[0027] Optionally, the constructing the intelligent analysis model for the behavior of the operator includes: obtaining a target image containing multi-category features to form an image data set; building an initial lightweight model for multi-category recognition, setting the hyperparameters for the training of the lightweight model, and performing model training iteration on the initial lightweight model according to the image data set until convergence; and constructing a lightweight model for real-time detection of multi-category features.

[0028] Optionally, the building of the initial lightweight model for multi-category recognition further includes clustering teacher behavior sub-models according to the types of the target image features, coupling the teacher behavior sub-models based on the features of the behavior to be recognized, performing parallel processing on the target image based on the coupled multiple teacher behavior sub-models, and constructing multiple student behavior sub-models based on the multiple teacher behavior sub-models; and simultaneously outputting multiple behavior recognition results through the multiple student behavior sub-models.

[0029] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses an intelligent monitoring system for chemical enterprise operation and construction, which has the following beneficial effects:

[0030] The present invention provides an intelligent monitoring system for chemical enterprise operation and construction, including: Equipment layer: including special operation intelligent guardian equipment and special operation personnel positioning equipment, which are used to provide data support for the algorithm layer and the platform layer; Algorithm layer: used to construct an intelligent analysis model for the behavior of operation personnel and a real-time positioning and supervision model for operation personnel; Platform layer: including a video management and AI intelligent analysis system and a special operation permission and operation process management system. The video management and AI intelligent analysis system is used to store and manage the video data collected during the operation, and at the same time call the models in the algorithm layer for data analysis. The special operation permission and operation process management system is used to provide the full-process management functions of special operation application, process approval, execution monitoring, and post-event analysis. Through the special operation intelligent guardian equipment that integrates gas continuous sampling, real-time video monitoring, and intelligent analysis of on-site personnel behavior, and through the software platform to integrate and analyze the real-time monitoring data, the present invention can not only effectively replace some traditional manual monitoring work, but also significantly improve the safety supervision efficiency and management level of special operations. With the intelligent guardian equipment as the core, integrating technologies such as artificial intelligence and personnel positioning, the present invention explores and studies the intelligent supervision technology for special operations in chemical enterprises. At the same time, combined with the software platform, it realizes the whole-process permission management of operations, video supervision of on-site operations, continuous sampling and analysis of toxic and harmful gases, and uses technologies such as visual analysis to accurately identify and warn of non-standard behaviors and unsafe scenarios during the whole operation process. Application practice shows that this research can provide a feasible and referenceable solution for the safety supervision of special operations in chemical enterprises. Description of the Drawings

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.

[0032] Figure 1 It is a structural framework diagram of an intelligent monitoring system for chemical enterprise operation and construction provided by the present invention.

[0033] Figure 2 It is a structural framework diagram of the special operation intelligent guardian equipment provided by the present invention. Detailed Embodiments

[0034] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0035] An embodiment of the present invention discloses an intelligent monitoring system for chemical enterprise operation and construction, as Figure 1 shown, including:

[0036] Device layer: including special operation intelligent guardian devices and special operation personnel positioning devices, used to provide data support for the algorithm layer and the platform layer;

[0037] Algorithm layer: used to construct an intelligent analysis model for the behavior of operation personnel and a real-time positioning and supervision model for operation personnel;

[0038] Platform layer: including a video management and AI intelligent analysis system and a special operation permission and operation process management system. The video management and AI intelligent analysis system is used to store and manage the video data collected during the operation process, and at the same time call the models in the algorithm layer for data analysis. The special operation permission and operation process management system is used to provide the full-process management functions of special operation application, process approval, execution monitoring, and post-event analysis.

[0039] Further, the special operation intelligent guardian device includes: an explosion-proof shell, a main control board, a pump suction system, a video acquisition system, a battery charging management module, a wireless communication module, an audible and visual alarm module, a digital display module, and a mobile terminal Bluetooth configuration module; the main control board, the pump suction system, the battery charging management module, the wireless communication module, the audible and visual alarm module, the digital display module, and the mobile terminal Bluetooth configuration module are placed inside the explosion-proof shell, and the main control board is respectively signal-connected to the pump suction system, the video acquisition system, the battery charging management module, the wireless communication module, the audible and visual alarm module, the digital display module, and the mobile terminal Bluetooth configuration module.

[0040] Further, the pump suction system includes an air inlet, an air outlet, an air pump, and a gas sensor. The air inlet, the gas sensor, the air pump, and the air outlet are connected in sequence, and the gas sensor is signal-connected to the main control board.

[0041] Further, the video management and AI intelligent analysis system includes: a device management module, an algorithm configuration module, a real-time monitoring module, a historical video module, and an intelligent analysis module;

[0042] The device management module: includes hierarchical and classification management of intelligent guardian devices, maintains the basic information and bound power-on and power-off information of the devices, and provides an overview of the device status;

[0043] The algorithm configuration module: used for configuring and managing the intelligent analysis algorithm for operators' behaviors;

[0044] The real-time monitoring module: provides a split-screen display function, and is used for providing multiple real-time monitoring images transmitted by intelligent guardian devices and fixed cameras to users simultaneously;

[0045] The historical video module: used for storing and managing the video data collected by intelligent guardian devices and fixed cameras during the operation process, and supports video playback at the same time;

[0046] The intelligent analysis module: used for performing real-time analysis on the behaviors of personnel at the operation site based on the personnel behavior recognition algorithm, generating warning information when abnormal situations are found, and feeding it back to the intelligent guardian devices for operations and the special operation permit and operation process management system.

[0047] Furthermore, the special operation permit and operation process management system includes: a permit management module, an operation implementation module, an operation monitoring module, an operation alarm module, and a statistical analysis module;

[0048] The permit management module: used for providing functions such as operation application, safety measure confirmation, sampling analysis, and approval of operation permits. After the operation application, the system automatically pushes the task to the relevant personnel for measure confirmation, sampling analysis, and approval according to the process. The relevant personnel perform corresponding operations and upload the results through the special operation mobile terminal. At the same time, the system will automatically associate the operation-related information, generate an electronic operation ticket according to the standard operation ticket template, and support the preview and download of the operation ticket;

[0049] The operation implementation module: used for providing functions such as operation start, pause, resume, end, and safety disclosure for operation implementation management. At the same time, when the operation starts, it supports scanning the code to bind the intelligent guardian device and supports viewing the real-time alarm information during the operation process;

[0050] The operation monitoring module: supports managers to remotely view the real-time information of the operation site;

[0051] The operation alarm module: used for integrating various warning analysis algorithms for alarm analysis, notifying the relevant personnel of the alarm information, and supporting tracking the progress of alarm handling;

[0052] The statistical analysis module: used for providing functions of operation and alarm statistical analysis, and summarizing and analyzing the on-site operation tickets and alarm situations according to the dimensions of time, operation ticket type, applying unit, and alarm type.

[0053] Furthermore, the real-time positioning and supervision model for operators includes: inputting real-time monitoring images into an object detection algorithm to detect the human bodies in the images and mark the position information; judging whether an employee is absent from the post according to the human bodies and the position information. If no human body target is detected in a certain employee's area for more than a preset time threshold, it is determined that the employee has left the post; if the number of human bodies exceeds the preset number threshold when the human bodies are marked by the object detection algorithm in a certain area, it is determined that the area is overcrowded.

[0054] Furthermore, it also includes: when a human body target is detected in an area, identifying the human bodies and markers in the area; determining the correlation between the human bodies and the markers; judging whether the human bodies in the target area are target objects according to the correlation; determining the off-duty status of the employees in the area according to the judgment result.

[0055] Furthermore, the determination of the correlation between the human bodies and the markers includes: determining the human body areas occupied by the human bodies, and determining the marker areas occupied by the markers; determining the coincidence situation between the human body areas and the marker areas, where the coincidence situation is the ratio of the area of the overlapping area between the human body areas and the marker areas to the area of the marker areas; determining the correlation between the human bodies and the markers based on the coincidence situation.

[0056] Furthermore, the construction of the intelligent analysis model for operators' behaviors includes: obtaining target images with multi-category features to form an image data set; building an initial lightweight multi-category recognition model and setting the hyperparameters for lightweight model training, and performing model training iteration on the initial lightweight model according to the image data set until convergence; constructing a lightweight model for real-time detection of multi-category features.

[0057] Furthermore, the construction of the initial lightweight multi-category recognition model also includes clustering teacher behavior sub-models according to the types of target image features, coupling the teacher behavior sub-models based on the features of the behaviors to be recognized, performing parallel processing on the target images based on the multiple coupled teacher behavior sub-models, and constructing multiple student behavior sub-models based on the multiple teacher behavior sub-models; simultaneously outputting multiple behavior recognition results through the multiple student behavior sub-models.

[0058] In the specific implementation, an intelligent monitoring system for the operation and construction of chemical enterprises constructs an efficient and intelligent supervision system for the special operation scenarios of chemical enterprises. In the stages of job ticket submission and approval, through the job permit management module, the online process and standardized management of job tickets are realized, thus greatly improving the efficiency and transparency of job approval and execution. During the implementation of the operation, the intelligent guardian device collects the on-site video images and the concentration of toxic and harmful gases in real time. At the same time, through the intelligent analysis technology of the behavior of operating personnel and the real-time positioning and supervision technology of personnel, the abnormal behaviors and potential safety hazards of operating personnel are captured and analyzed in real time. The monitoring and early warning information can be integrated into the special operation supervision software platform, which is convenient for safety supervisors and managers to insight into the safety status during the operation process in real time, and to discover and respond to potential safety risks in the first time. In terms of design, this embodiment adopts a clear layered architecture to ensure the flexibility and scalability of the system. The overall architecture is as Figure 1 shown.

[0059] Device layer: The device layer includes the special operation intelligent guardian device and the special operation personnel positioning device. Its main function is to collect the on-site video materials, gas concentration data during the operation in real time, and the personnel behavior recognition information obtained by means of AI technology. At the same time, the position information of operating personnel is accurately obtained. The core function of the device layer is to provide basic data support for the algorithm layer and the platform layer, which is the premise for the operation of the entire intelligent supervision system.

[0060] Algorithm layer: The algorithm layer integrates a variety of algorithms specifically for the intelligent analysis of the behavior of on-site operating personnel, including visual analysis algorithms such as flame recognition, safety helmet recognition, work clothes recognition, personnel falling recognition, reflective vest recognition, mobile phone playing recognition, fire extinguisher placement recognition, double hook recognition, and personnel protection equipment recognition, as well as personnel real-time positioning and supervision algorithms such as personnel leaving the post, breaking in, and overcrowding in the area.

[0061] Platform layer: The platform layer consists of two core parts: the video management and AI intelligent analysis system and the special operation permit and operation process management system. The video management and AI intelligent analysis system is responsible for the efficient storage and management of the video data collected during the operation. At the same time, it can call the personnel behavior intelligent analysis algorithm of the algorithm layer for analysis and feedback the analysis results to the special operation intelligent guardian device and the software system in time. The special operation permit and operation process management system covers multiple key functional modules such as permit management, operation implementation, operation monitoring, operation alarm, and statistical analysis, covering the full-process management functions of special operation application, process approval, execution monitoring, and post-event analysis, ensuring the whole-process monitoring of the operation and effectively improving the safety and management efficiency of the operation.

[0062] In the specific implementation, the special operation intelligent guardian device is a highly integrated all-in-one product. Its R & D process focuses on convenience, integration, and intelligence, aiming to ensure the safety monitoring and efficient management of the operation site through the integration of precise design and advanced technical means.

[0063] Its core components include the device control center, toxic and harmful gas sensors and pump suction system, explosion-proof infrared video acquisition module, AI-based intelligent analysis module for on-site operator behavior, mobile Bluetooth configuration module, digital display and voice-light alarm module, battery charging management module, 5G wireless communication module, explosion-proof shell and packaging, etc. Among them, the device control center, as the core brain, is responsible for integrating and coordinating each functional module to achieve unified data processing and instruction issuance; the toxic and harmful gas sensors and pump suction system can accurately monitor the concentration of harmful gases in the operation environment to achieve real-time monitoring and early warning; the explosion-proof infrared video acquisition module provides high-definition visual monitoring, and at the same time, combined with the AI-based intelligent analysis technology for on-site operator behavior, it can automatically identify illegal operations and potential risks; the mobile Bluetooth configuration interface is used for rapid configuration and wireless connection of the device, improving the flexibility and convenience of on-site deployment; the digital display and voice-light alarm module immediately feedbacks the monitoring results to ensure that information is quickly conveyed to on-site operators in case of emergencies; the battery charging management module is responsible for optimizing the battery usage efficiency to ensure the continuous and stable operation of the device during long-term monitoring tasks; the 5G wireless communication module enables high-speed data transmission to ensure the real-time nature of remote monitoring and data synchronization; in addition, an explosion-proof shell and professional packaging are adopted to ensure the stable operation of the device in harsh environments. The overall structure of the device is as Figure 2 shown, and the specific components include:

[0064] Device control center: The device control center adopts a high-performance industrial-grade 32-bit communication processor and an industrial-grade wireless module, and uses an embedded real-time operating system as its software support platform to ensure the high efficiency and stability of system operation. This control center provides a rich interface design, enabling seamless information interaction and instruction transfer with other functional modules, thereby collaborating to complete the overall functions of the device. This high-performance design architecture enables the device to efficiently process multi-task data acquisition, analysis, and transmission requirements, and lays a solid foundation for the scalability and intelligence of the entire system.

[0065] Toxic and harmful gas sensors and pump suction system: The toxic and harmful gas sensors and pump suction module adopt a pump suction gas collection method. The CPU of the device control center precisely controls the power supply of the air pump to ensure that the air pump can continuously and stably extract gas from the on-site environment and transport it to the toxic and harmful gas sensors for real-time sample collection. The toxic and harmful gas sensors can accurately detect H2S, CO, O2, NH3, NO xThe concentrations of six gases, namely (nitrogen oxides) and EX (flammable gas). The CPU of the equipment control center analyzes the collected gas concentration data to judge in real time whether it exceeds the preset safety threshold. Once it is found that the concentration of a certain gas exceeds the standard, the system will immediately activate a dual warning mechanism: on the one hand, the alarm information will be uploaded to the special operation software platform through the wireless communication module, so that remote managers can timely grasp the on-site situation; on the other hand, it will trigger the on-site sound and light alarm and voice prompt module to quickly remind the operators to take countermeasures and reduce safety risks. The air pump adopts a diaphragm design based on the principle of positive displacement pump, is equipped with a high-performance brushless motor and a high-quality engineering plastic shell, and has the characteristics of efficient operation, stable performance, durability and reliability. It can work normally for a long time in a complex environment, providing a strong guarantee for the stability and accuracy of gas monitoring.

[0066] Explosion-proof infrared video acquisition module: The equipment integrates an explosion-proof infrared camera, which has the safety and efficiency to adapt to complex working environments. The camera uses wireless communication technology and is built-in with a Wi-Fi module and a long-lasting power supply battery, enabling it to continuously collect and transmit on-site images at special operation sites. The camera provides two convenient installation methods: bracket fixation and strong magnetic suction base, which facilitate the operators to flexibly adjust the equipment position according to the actual working conditions. The on-site video data collected by the camera will be pushed to the video management and AI intelligent analysis system through the wireless network. On the one hand, it will conduct real-time in-depth analysis to accurately capture abnormal details in the images; on the other hand, it will complete the operation of storing and leaving traces, providing complete information for subsequent backtracking and verification. With powerful computing power, it can screen each frame of the video content, quickly identify abnormal conditions such as illegal operations and environmental anomalies, and give early warning reminders in time to effectively ensure the orderly and safe progress of special operations.

[0067] Mobile Bluetooth configuration module: To facilitate the configuration of equipment parameters, a Bluetooth communication module is integrated inside the equipment. Through the mobile phone, information such as the threshold range of each gas concentration and the system time can be set, and the real-time sampling value of the on-site gas concentration, battery capacity, etc. can also be queried, thus improving the convenience of equipment maintenance.

[0068] Digital display and voice sound and light alarm module: The equipment is equipped with a liquid crystal display, which can visually display the concentrations of six gases, namely H2S, CO, O2, NH3, nitrogen oxides, and EX flammable gas, the real-time time, the remaining power of the host, and the remaining power of the camera battery on-site, facilitating on-site operators to timely read the detection results. In addition, the equipment also integrates a sound and light alarm system. When the detected gas concentration exceeds the preset threshold or specific abnormal human behaviors are identified, the system will automatically trigger the sound and light alarm and play corresponding voice prompts according to different alarm types to ensure the timeliness of safety warnings. X The concentrations of six gases, namely nitrogen oxides and EX flammable gas, the real-time time, the remaining power of the host, and the remaining power of the camera battery, facilitating on-site operators to timely read the detection results. In addition, the equipment also integrates a sound and light alarm system. When the detected gas concentration exceeds the preset threshold or specific abnormal human behaviors are identified, the system will automatically trigger the sound and light alarm and play corresponding voice prompts according to different alarm types to ensure the timeliness of safety warnings.

[0069] Battery charging management module: The equipment battery charging management includes two parts: the main equipment battery and the video camera battery. The battery uses a large-capacity rechargeable battery of 12V / 30AH. After the battery is fully charged, it can meet the needs of continuous work for 8-10 hours to ensure operation continuity.

[0070] 5G wireless communication module: The device is integrated with a 5G communication module. The gas concentration information and monitoring video collected on site can be transmitted to the management platform in real time through the 5G network. The device supports a variety of network protocols, including PPP, TCP, UDP, ICMP, HTTP, MQTT, etc., and is compatible with multiple SOCKET interface standards, with good network adaptability. The communication protocol follows the standard MODBUS 485 method and reads information according to function codes and related instructions. The device flow card includes two parts: video card flow and gas concentration collection upload flow. Both use annual package services to ensure unlimited data upload during continuous operation of the device.

[0071] Explosion-proof housing and packaging: The device host, portable camera and bracket are embedded in custom-molded foam in the outer packaging box to ensure stability and protection during transportation and storage. The outer packaging box is customized with all-aluminum-magnesium alloy design, which is sturdy and durable. It is also equipped with an external pull rod and universal wheels, which makes it easy for users to transport the equipment from the management office to the site of use, improving the portability and ease of operation of the equipment. The main equipment box is customized with carbon steel welding and has passed the explosion-proof certification, which can be used safely in complex and dangerous environments. There are two explosion-proof buttons on the device panel, namely the power switch and the alarm cancellation button (used to manually stop the alarm after the sound and light alarm is triggered). There is a handle on the back of the box for easy carrying; the bottom is equipped with rubber non-slip fixed feet to ensure that the equipment can be placed stably on site and adapt to a variety of usage environments.

[0072] In the specific implementation, an intelligent analysis technology for the behavior of operators based on artificial intelligence is proposed. Due to the complexity of the on-site environment and the unpredictability of the behavior of operators, it is difficult for traditional video surveillance means to achieve real-time monitoring and intelligent analysis of the behavior of special personnel, and there are problems such as a large workload of manual discrimination, easy misjudgment and missed judgment. Using artificial intelligence technology to automatically identify abnormal behavior of personnel during the production operation process has become an important development direction. This embodiment is based on deep learning technology, and a neural network model is constructed to train a large number of labeled on-site image or video data to achieve accurate recognition and analysis of various target objects. During the training process, a large amount of relevant data is first collected and preprocessed, including images or videos of various target objects. Then, a convolutional neural network (CNN) is used to extract features and classify and learn the preprocessed data. Through continuous iterative training, an accurate recognition ability for target objects is formed. During the operation process, real-time video images are captured by a high-definition camera carried by a special operation intelligent guardian device, and the algorithm model can automatically identify and analyze target objects in the image content, so as to automatically identify unsafe behaviors, abnormal actions and potential safety hazards of personnel.

[0073] In the specific implementation, the real-time positioning and supervision model of the operator specifically includes: identifying the position of the human body target and the corresponding confidence level from the image intercepted frame through the Yolov8 object detection network; and screening according to the confidence level threshold to remove the results with lower confidence levels to obtain a preliminary prediction result; after the Yolov8 object detection network detects the human body target, the human body target is intercepted and a Resnet50 secondary classification network is used for the first round of screening to screen out some misidentified targets; for the human body targets screened by Resnet50, another round of screening is carried out. First, Gaussian smoothing filtering is used to eliminate Gaussian noise, then the canny edge detection algorithm is used to obtain its edge features, and finally the Hough line detection algorithm is used to detect the number of lines in the edge features; if the number of lines is less than a preset threshold, then these targets are further screened out; this step can screen out some non-human targets with more curves and fewer lines; after obtaining the final personnel target and work position information, the off-duty determination is made according to their mutual position information; off-duty determination: if no employee target is detected at the work position, it is determined as off-duty, and if the number of detected human bodies exceeds the preset number threshold, then it is determined that the area is overcrowded.

[0074] In the specific implementation manner, it further includes: identifying a regional image through a pre-trained object detection model to obtain the human body and markers in the regional image, determining the correlation between the human body and the markers, determining the human body area occupied by the human body, and determining the marker area occupied by the markers; determining the matching situation between the human body area and the marker area, where the matching situation is the ratio of the area of the overlapping area between the human body area and the marker area to the area of the marker area; based on the matching situation, determining the correlation between the human body and the markers. According to the correlation, determining whether the human body in the target area is a target object, and obtaining a preset correlation threshold between the target object and the markers; if the correlation is greater than the preset correlation threshold, determining that the human body in the area is a target object; if the correlation is not greater than the preset correlation threshold, determining that the human body in the area is not a target object; according to the judgment result, determining the off-duty status of the employees in the area, and when the human body in the area is not a target object, counting the off-duty time of the target object; if it is detected that the off-duty time exceeds the preset specified time, determining that the target object is in an off-duty status, and if there is no human body and / or on-duty markers in the regional image, determining that the target object is in an off-duty status.

[0075] In the specific implementation manner, a real-time positioning and supervision technology for on-site operators is proposed. In a complex on-site operation environment, it is crucial to ensure the safety of operators and the efficient management of operation order. In this embodiment, the Bluetooth device carried by the operator communicates instantaneously with the Bluetooth beacon to obtain the real-time position of the operator. At the same time, based on the real-time position of the personnel, a comprehensive personnel behavior monitoring algorithm is studied and implemented. This algorithm can intelligently identify the off-duty situation of the guardians, the behavior of unauthorized personnel breaking in, and the overcrowding situation in the operation area, thus ensuring the efficient and accurate supervision and management of on-site personnel and providing strong technical support for the safety and order of on-site operations.

[0076] In the specific implementation manner, the construction of the intelligent analysis model for operator behavior specifically includes:

[0077] Obtaining a target image containing multi-category features to form an image data set;

[0078] Building an initial multi-category recognition lightweight model as the teacher model and setting the hyperparameters for lightweight model training. Among them, the initial lightweight model includes multiple lightweight backbone networks and neck networks, and the lightweight backbone networks are used to extract multi-scale image features that fuse position information;

[0079] Performing data augmentation on the image to obtain a preprocessed image; iterating the following steps until convergence:

[0080] The preprocessed image is input into the initial first lightweight model. Multiple compressed images are obtained through the Focus unit, and multi-channel connection and convolution are performed on the multiple compressed images to form n 11 feature maps; The n 11 feature maps are input into the first MobileNetV3 module for downsampling to obtain the n 12 feature maps; The n 12 feature maps are sequentially input into the second MobileNetV3 module after passing through the HDC unit, and the output is the n 13 feature maps; The preprocessed image is input into the initial nth lightweight model. Multiple compressed images are obtained through the Focus unit, and multi-channel connection and convolution are performed on the multiple compressed images to form the nth i1 feature maps; The n i1 feature maps are input into the first MobileNetV3 module for downsampling to obtain the n i2 feature maps; The n i2 feature maps are sequentially input into the second MobileNetV3 module after passing through the HDC unit, and the output is the n i3 feature maps; The n 11 、...、n i3 feature maps are input into the neck network for feature information fusion;

[0081] A student model is established according to multiple lightweight backbone networks and the neck network. The teacher model trains the student model through knowledge distillation to obtain a lightweight student model. A head detection network is constructed under the student model, and the result of object detection is given through the detection network. The error value is calculated by comparing the predicted value and the true value of the object detection, and the training process is iterated;

[0082] Based on the comparison between the detection result and the true value, a loss function is calculated. Specifically, the loss function is expressed as: LOSS = LOSS cls + LOSS bbox + LOSS obj + LOSS dw1 + LOSS dw2 , where Loss is the total loss of the initial lightweight model, Loss cls is the classification loss, Loss bbox is the rectangular box regression loss, Loss obj is the confidence loss, LOSS dw1 is the lightweight backbone network distillation loss, LOSS dw2 is the neck network distillation loss.

[0083] In the specific implementation, the neck network selects FPN-PAN in the Yolov5 structure. This network uses multiple feature map samplings, feature map cascades, and the C3 structure (CSP Bottleneck with 3 Convolutions) to achieve information fusion of multi-scale image features, thereby enhancing the network's learning ability. The detection network corresponding to the FPN-PAN neck network selects the head detection network of Yolov5. Among them, the head detection network can be composed of n prediction detection heads, and the n prediction detection heads are respectively connected to the FPN-PAN neck network. The head detection network corresponds to the object detection results of the neck network at different scales, thus forming a complete initial lightweight model.

[0084] In the specific implementation, the construction of the initial lightweight model for multi-category recognition specifically includes: clustering teacher behavior sub-models according to the types of target image features. Among them, the behaviors to be recognized include flame recognition, safety helmet recognition, work clothes recognition, personnel falling to the ground recognition, reflective vest recognition, mobile phone playing recognition, fire extinguisher placement recognition, double hook recognition, and personnel protection equipment recognition. Extract features from the target image, extract the feature types required for the behaviors to be recognized from the extracted features, plan to set teacher behavior sub-models with the number of behavior types to be recognized, merge teacher behavior sub-models with similar feature types according to the required feature types to obtain multiple comprehensive teacher behavior sub-models, couple multiple comprehensive teacher behavior sub-models with multiple teacher behavior sub-models to perform parallel processing on the target image, construct multiple student behavior sub-models under parallel processing, and simultaneously output multiple behavior recognition results through multiple student behavior sub-models.

[0085] In the specific implementation, the video management and AI intelligent analysis system specifically includes: device management, algorithm configuration, real-time monitoring, historical videos, and intelligent analysis, aiming to provide a solid storage infrastructure and intelligent algorithm support for the special operation permit and operation process management system.

[0086] (1) Device management: It includes hierarchical and classified management of intelligent guardian devices, maintains the basic information and bound power-on and power-off information of the devices, provides an overview of the device status, and provides services for the efficient organization and convenient maintenance of the devices.

[0087] (2) Algorithm configuration: Realize flexible configuration management of the intelligent analysis algorithm for the behaviors of on-site operators, including flame recognition, safety helmet recognition, work clothes recognition, personnel falling to the ground recognition, reflective vest recognition, fire extinguisher placement recognition, double hook recognition, and personnel protection equipment recognition, etc.

[0088] (3) Real-time monitoring: Provide a split-screen display function, which is convenient for users to view the real-time monitoring images transmitted by multiple intelligent guardian devices and fixed cameras at the same time.

[0089] (4) Historical videos: Store and manage the video data collected by the intelligent guardian device and fixed cameras during the operation process. At the same time, video playback is supported to ensure the traceability of the entire operation process.

[0090] (5) Intelligent analysis: Based on the personnel behavior recognition algorithm, conduct real-time analysis on the behaviors of personnel at the operation site. Once abnormal situations are detected, warning information will be generated immediately and fed back to the intelligent guardian device for operations and the special operation permit and operation process management system in real time.

[0091] In the specific implementation manner, the special operation permit and operation process management system specifically includes: permit management, operation implementation, operation monitoring, alarm management, and statistical analysis. Its main purpose is to realize the digital and standardized management of the entire process from application, approval to acceptance and archiving of operations. At the same time, based on the intelligent guardian device, continuous and uninterrupted gas monitoring, video monitoring, and intelligent analysis and alarm are realized, and real-time monitoring and risk warning of the operation process are comprehensively achieved.

[0092] (1) Permit management: The system provides functions for operation permit management such as operation application, safety measure confirmation, sampling analysis, and approval. After the operation application, the system automatically pushes the task to relevant personnel such as measure confirmation, sampling analysis, and approval according to the process. Relevant personnel can perform corresponding operations and upload the results through the special operation mobile terminal. At the same time, the system will automatically associate the operation-related information, generate an electronic operation ticket according to the standard operation ticket template, and support the preview and download of the operation ticket.

[0093] (2) Operation implementation: The system provides functions for operation implementation management such as operation start, pause, resume, end, and safety disclosure. At the same time, when the operation starts, it supports scanning the code to bind the intelligent guardian device, and supports viewing the real-time alarm information during the operation process for quick handling.

[0094] (3) Operation monitoring: Managers can remotely view the real-time information of the operation site through the system, including the real-time video of the bound intelligent guardian device, the monitoring video of the fixed cameras in the operation area, the real-time concentration of toxic and harmful gases (H2S, CO, O2, NH3, NO X nitrogen oxides, EX combustible gas) collected by the intelligent guardian device and the historical curve, as well as the real-time alarm information of the operation site, which is convenient for managers to effectively grasp the dynamics of the operation site and ensure the safety of the operation environment.

[0095] (4) Alarm Management: Integrate multiple early warning analysis algorithms, including intelligent analysis alarms for personnel behavior and personnel location alarms. Among them, the intelligent analysis module for personnel behavior is mainly responsible for receiving alarm pushes from the video management and AI intelligent analysis systems, including alarm information such as flames, safety helmets, work clothes, personnel falling to the ground, reflective vests, fire extinguisher placement, double hooks, and personnel protective equipment. Personnel location analysis alarms include alarm information such as personnel leaving their posts, personnel breaking in, and overcrowding in areas. The system will notify relevant personnel of the alarm information through various methods such as text messages, DingTalk, and audible and visual alerts on on-site intelligent guardian devices, and support tracking the progress of alarm handling.

[0096] (5) Statistical Analysis: The system provides functions for statistical analysis of operations and alarms, and can summarize and analyze on-site work tickets and alarm situations according to multiple dimensions such as time, work ticket type, applying unit, and alarm type, helping management comprehensively grasp the operation status and on-site risks, and providing data support for the advance planning and implementation of preventive measures.

[0097] The present invention takes the intelligent guardian of special operations as the core, and comprehensively expounds the intelligent supervision technology and software platform functions of special operations of chemical enterprises. The research results can realize comprehensive supervision of the whole process of special operations and significantly improve the safety level of operations. In the future, the intelligent analysis algorithm of the behavior of on-site operators can be further expanded to achieve more comprehensive and accurate supervision and identification capabilities, and further strengthen the role of intelligent supervision technology in ensuring operation safety. The special operation intelligent supervision technology and software platform developed by the present invention have been officially put into use in Shanxi Zhongmei Pingshuo Energy Chemical Co., Ltd. In the operation test, under full load working conditions (continuous triggering and response of sound, light and voice alarms), a single charge can maintain efficient continuous operation for 26.8 hours. Considering that special operations usually take 8 to 10 hours a day, the equipment can be used uninterruptedly for 2 to 3 days, which can fully meet the needs of continuous online monitoring. In addition, since the launch of the intelligent supervision technology and software platform for special operations in June 2024, it has provided enterprises with a total of more than 3,000 special operations safety supervision services, and the platform has identified more than 1,000 alarm events, and can push them to the relevant responsible persons in a timely and accurate manner. By comparing and analyzing the data before and after the platform was launched, it was found that after the platform was put into use, the average signature circulation cycle was significantly shortened, from about 1 hour to 24 minutes, and the overall efficiency was greatly improved. In addition, in the initial application stage of the platform, due to the irregular filling of paper work tickets, there were many violations in the filling and approval process of work tickets. After a period of trial operation, the violations have basically disappeared. In the actual implementation of the operation, the violations of personnel behavior have also been significantly reduced, and there are basically no obvious violations at present. Through the efficient monitoring and early warning mechanism of the platform, it has not only significantly improved the ability of enterprises in the safety management of special operations, but also played a key role in discovering violations of operation and identifying unsafe conditions of personnel environment. Through real-time monitoring and intelligent analysis, the platform can timely reveal potential safety hazards, while effectively supervising and restraining operators.

[0098] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0099] The foregoing description of the disclosed embodiments enables those skilled in the art to practice or use the present invention. Various modifications to these embodiments will be readily 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. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent monitoring system for chemical enterprise operation and construction, characterized in that It includes: Device layer: It includes special operation intelligent guardian devices and special operation personnel positioning devices, which are used to provide data support for the algorithm layer and the platform layer; Algorithm layer: It is used to build an intelligent analysis model for the behavior of operation personnel and a real-time positioning and supervision model for operation personnel; Platform layer: It includes a video management and AI intelligent analysis system and a special operation permit and operation process management system. The video management and AI intelligent analysis system is used to store and manage the video data collected during the operation process, and at the same time call the models of the algorithm layer for data analysis. The special operation permit and operation process management system is used to provide the full-process management functions of special operation application, process approval, execution monitoring, and post-event analysis.

2. The intelligent monitoring system for chemical enterprise operation construction according to claim 1, characterized in that, The special operation intelligent guardian device includes: an explosion-proof shell, a main control board, a pump suction system, a video acquisition system, a battery charging management module, a wireless communication module, an audible and visual alarm module, a digital display module, and a mobile Bluetooth configuration module; the main control board, the pump suction system, the battery charging management module, the wireless communication module, the audible and visual alarm module, the digital display module, and the mobile Bluetooth configuration module are placed inside the explosion-proof shell, and the main control board is respectively signal-connected to the pump suction system, the video acquisition system, the battery charging management module, the wireless communication module, the audible and visual alarm module, the digital display module, and the mobile Bluetooth configuration module.

3. The intelligent monitoring system for chemical enterprise operation construction according to claim 2, wherein, The pump suction system includes an air inlet, an air outlet, an air pump, and a gas sensor. The air inlet, the gas sensor, the air pump, and the air outlet are connected in sequence, and the gas sensor is signal-connected to the main control board.

4. An intelligent monitoring system for chemical enterprise operation construction according to claim 1, characterized in that, The video management and AI intelligent analysis system includes: a device management module, an algorithm configuration module, a real-time monitoring module, a historical video module, and an intelligent analysis module; The device management module: It includes the hierarchical and classification management of intelligent guardian devices, maintains the basic information of the devices and the bound power-on and power-off information, and provides an overview of the device status; The algorithm configuration module: It is used to configure and manage the intelligent analysis algorithm for the behavior of operation personnel; The real-time monitoring module: It provides a split-screen display function, which is used to provide the user with real-time monitoring pictures transmitted by multiple intelligent guardian devices and fixed cameras at the same time; The historical video module: It is used to store and manage the video data collected by the intelligent guardian device and the fixed camera during the operation process, and at the same time supports video playback; The intelligent analysis module: It is used to perform real-time analysis on the behavior of personnel at the operation site based on the personnel behavior recognition algorithm, generate warning information when abnormal situations are found, and feedback it to the operation intelligent guardian device and the special operation permit and operation process management system.

5. An intelligent monitoring system for chemical enterprise operation construction according to claim 1, characterized in that, The special operation permit and operation process management system includes: a permit management module, an operation implementation module, an operation monitoring module, an operation alarm module, and a statistical analysis module; The permission management module: It is used to provide functions for job application, safety measure confirmation, sampling analysis, and approval of job permits. After a job application, the system automatically pushes tasks to relevant personnel for measure confirmation, sampling analysis, and approval according to the process. Relevant personnel perform corresponding operations and upload results through the special operation mobile terminal. At the same time, the system will automatically associate job-related information, generate an electronic job ticket according to the standard job ticket template, and support preview and download of the job ticket. The job implementation module: It is used to provide functions for job start, pause, resume, end, and safety disclosure job implementation management. At the same time, when the job starts, it supports scanning the code to bind the intelligent guardian device and viewing real-time alarm information during the job process. The job monitoring module: It supports managers to remotely view real-time information of the job site. The job alarm module: It is used to integrate various early warning analysis algorithms for alarm analysis, notify relevant personnel of the alarm information, and support tracking the progress of alarm handling. The statistical analysis module: It is used to provide functions for job and alarm statistical analysis, and summarize and analyze the on-site job tickets and alarm situations according to the dimensions of time, job ticket type, applying unit, and alarm type.

6. The intelligent monitoring system for chemical enterprise operation construction according to claim 1, characterized in that, The real-time positioning and supervision model of job personnel includes: inputting the real-time monitoring image into the object detection algorithm to detect the human body in the image and mark the position information; judging whether an employee is off the post according to the human body and the position information. If no human body target is detected in a certain employee's area for more than a preset time threshold, then it is determined that the employee has left the post; if the number of human bodies exceeds the preset number threshold when the human body is marked in a certain area through the object detection algorithm, then it is determined that the area is overcrowded.

7. An intelligent monitoring system for chemical enterprise operation construction according to claim 6, characterized in that, It also includes: When a human body target is detected in the area, identifying the human body and the marker in the area; Determining the correlation between the human body and the marker; Judging whether the human body in the target area is the target object according to the correlation; Determining the off-the-post status of the area employees according to the judgment result.

8. An intelligent monitoring system for chemical enterprise operation construction according to claim 7, characterized in that, The determination of the correlation between the human body and the marker includes: determining the human body area occupied by the human body, and determining the marker area occupied by the marker; determining the coincidence situation between the human body area and the marker area, where the coincidence situation is the ratio of the area of the overlapping area between the human body area and the marker area to the area of the marker area; determining the correlation between the human body and the marker based on the coincidence situation.

9. An intelligent monitoring system for chemical enterprise operation construction according to claim 1, characterized in that, The construction of the intelligent analysis model for job personnel behavior includes: obtaining target images with multi-category features to form an image data set; building an initial lightweight model for multi-category recognition and setting hyperparameters for lightweight model training, and performing model training iteration on the initial lightweight model according to the image data set until convergence; constructing a lightweight model for real-time detection of multi-category features.

10. An intelligent monitoring system for chemical enterprise operation construction according to claim 9, characterized in that, The construction of the initial lightweight multi-class recognition model further includes clustering teacher behavior sub-models according to the types of target image features, coupling the teacher behavior sub-models based on the features of the behavior to be recognized, performing parallel processing on the target image based on multiple teacher behavior sub-models obtained after coupling, and constructing multiple student behavior sub-models based on the multiple teacher behavior sub-models; multiple behavior recognition results are simultaneously output through the multiple student behavior sub-models.

Images