Limited space operation potential safety hazard identification and early warning method and device

Abstract

The invention provides a limited space operation potential safety hazard identification early warning method and device, and the method comprises the steps: carrying out the fusion of images captured by image collection devices which are scattered at a plurality of positions of a limited space operation site, and obtaining a limited space operation panoramic image; inputting the confined space operation panoramic image into a trained deep learning model to identify personal protection equipment and on-site safety facilities in the confined space operation panoramic image, and obtaining an identification result; the identification result comprises the wearing condition of personal protection equipment and the setting condition of field safety facilities; configuring a trigger condition for generating an early warning signal, and generating a corresponding early warning signal based on an identification result; and sending the early warning signal to a mobile device or a control center through a plurality of channels. The working state of the limited space can be monitored in real time, potential safety hazards are automatically recognized, early warning is given out in time, safety accidents caused by human factors are remarkably reduced, and a safer and more reliable working environment is provided for working in the limited space.

Description

Technical Field

[0001] The present invention relates to the technical field of confined space operation safety monitoring, and in particular to a method and device for identifying and warning of safety hazards in confined space operations. Background Art

[0002] Confined spaces such as basements, tunnels, and closed pipes have small spaces and poor working environments, which pose great challenges to security monitoring. Maintenance and inspection operations in these confined spaces not only face physical problems such as poor air quality and insufficient lighting, but also face safety risks such as unclear escape routes and poor communications. However, most of the current monitoring of these environments still relies on traditional physical access and visual inspection methods, which are not only inefficient but also extremely risky and difficult to meet the needs of modern security management.

[0003] Traditional methods of confined space safety monitoring mainly rely on basic electronic monitoring equipment for manual inspection. However, these methods have many shortcomings. First, they cannot achieve all-weather real-time monitoring, and can only monitor during specific time periods, which cannot ensure safety at all times. Secondly, the data processing and response speed are slow, and potential safety hazards cannot be discovered and handled in a timely manner. Furthermore, these methods are highly dependent on the experience and judgment of the operator. Once the operator makes a mistake in judgment, it may lead to serious consequences.

[0004] What’s more serious is that manual inspections also have many limitations in these limited spaces. Due to the small space, it is difficult for inspectors to enter certain areas for inspection, resulting in incomplete inspection coverage. At the same time, the frequency of manual inspections is relatively low, and potential safety hazards cannot be discovered and dealt with in a timely manner. In addition, manual inspections are easily affected by subjective factors, resulting in reduced inspection quality. Summary of the invention

[0005] In view of the above problems, the embodiments of the present invention provide a method and device for identifying and warning safety hazards in confined space operations to solve the existing technical problems.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for identifying and warning safety hazards in confined space operations, the method comprising:

[0008] The images captured by the image acquisition devices scattered at multiple locations on the limited space operation site are fused to obtain a panoramic image of the limited space operation;

[0009] Inputting the confined space operation panoramic image into the trained deep learning model to identify the personal protective equipment and on-site safety facilities in the confined space operation panoramic image, and obtaining a recognition result; the recognition result includes the wearing condition of the personal protective equipment and the setting condition of the on-site safety facilities;

[0010] Configure the trigger conditions for generating warning signals, and generate corresponding warning signals based on the recognition results;

[0011] Send warning signals to mobile devices or control centers through multiple channels.

[0012] In one embodiment, the step of fusing images captured by image acquisition devices dispersed at multiple locations on a confined space operation site to obtain a panoramic image of the confined space operation includes:

[0013] The entire working area is captured by a panoramic camera installed at the entrance of the limited space to obtain a first image;

[0014] An infrared camera installed in a limited space is used to collect an image in a low-light environment to obtain a second image;

[0015] A local close-up image is collected by using a high-resolution camera installed at an operating position of the equipment in a limited space to obtain a third image;

[0016] The second image and the third image are fused with the first image to obtain a panoramic image of the limited space operation.

[0017] In one embodiment, the training method of the deep learning model includes:

[0018] Collect images of limited space work sites from different channels to obtain training image data;

[0019] Label the personal protective equipment and on-site safety facilities in the training image data to obtain a training data set; the personal protective equipment includes a safety helmet, a respirator and a safety belt, and the on-site safety facilities include fencing facilities, safety signs, warning signs, safety notice boards, forced air supply equipment, lighting equipment, intercom equipment, a tripod and a gas detection alarm.

[0020] The training data set is input into the deep learning model for training to obtain a trained deep learning model.

[0021] In one embodiment, inputting the training data set into the deep learning model for training to obtain a trained deep learning model includes:

[0022] The training data set is divided into training set, validation set and test set according to three ratios;

[0023] The training set is used to make the model learn the mapping relationship from input data to output data. During the learning process, the pre-trained weights on the training data set are used as the initial weights.

[0024] The performance of the model is evaluated using a validation set, and the hyperparameters are adjusted according to the performance evaluation results until the model performance reaches a predetermined effect, thereby obtaining an initial deep learning model; the hyperparameters include a learning rate and the number of iterations, and the learning rate gradually decreases during the training process;

[0025] The test set is used to evaluate the generalization performance of the initial deep learning model, and the initial deep learning model whose generalization performance meets the expected effect is selected as the trained deep learning model.

[0026] In one embodiment, the configuration of the triggering condition for generating the warning signal, generating the corresponding warning signal based on the recognition result includes:

[0027] Identify equipment warning pools based on personal protective equipment and all equipment included in the on-site safety facility;

[0028] A corresponding warning switch is configured for each personal protective equipment and on-site safety facility; if the corresponding warning switch is in the on state, the corresponding personal protective equipment or on-site safety facility is added to the equipment warning pool; if the corresponding warning switch is in the off state, the corresponding personal protective equipment or on-site safety facility is removed from the equipment warning pool;

[0029] Assign an identification identifier to each confined space operation;

[0030] Setting the opening and closing of each warning switch according to the identification requirements of the confined space operation corresponding to each identity identifier;

[0031] Generate warning information on wearing of personal protective equipment according to the wearing status of personal protective equipment and the opening and closing status of the warning switch;

[0032] Generate early warning information on the setting of on-site safety facilities based on the setting of on-site safety facilities and the opening and closing status of early warning switches.

[0033] In one embodiment, generating warning information about wearing of personal protective equipment according to the wearing condition of the personal protective equipment and the opening and closing condition of the warning switch includes:

[0034] Generate a person target frame based on the identified persons in the limited space;

[0035] Identify personal protective equipment within the personnel target frame and obtain the wearing status of the personal protective equipment;

[0036] When the warning switch is set to the on state and the corresponding personal protective equipment is not recognized, warning information about the wearing of the personal protective equipment is generated.

[0037] In one embodiment, the generating of the warning information of the on-site safety facility setting according to the setting of the on-site safety facility and the opening and closing of the warning switch includes:

[0038] A flag bit is configured for each device included in the on-site safety facility. When the corresponding device is identified, the corresponding flag bit is set to 1; when the corresponding device is not identified, the corresponding flag bit is set to 0;

[0039] Initialize the flag bits of all devices to 0 at the beginning of limited space operation;

[0040] When the warning switch is set to the on state and the corresponding on-site safety facility flag is 0, a warning message about the on-site safety facility setting is generated;

[0041] When the warning switch is set to the on state and the corresponding on-site safety facility flag changes from 0 to 1, the warning information about the on-site safety facility setting generated before this operation is eliminated.

[0042] In a second aspect, the present invention provides a device for identifying and warning safety hazards in confined space operations, characterized in that the device comprises:

[0043] Panoramic image acquisition module: used to fuse images captured by image acquisition devices scattered at multiple locations on the limited space operation site to obtain a panoramic image of the limited space operation;

[0044] Recognition result acquisition module: used to input the confined space operation panoramic image into the trained deep learning model to identify the personal protective equipment and on-site safety facilities in the confined space operation panoramic image, and obtain the recognition result; the recognition result includes the wearing condition of the personal protective equipment and the setting condition of the on-site safety facilities;

[0045] Warning information generation module: used to configure the trigger conditions for generating warning signals and generate corresponding warning signals based on the recognition results;

[0046] Warning information sending module: used to send warning signals to mobile devices or control centers through multiple channels.

[0047] In one embodiment, the panoramic image acquisition module is specifically used for:

[0048] The entire working area is captured by a panoramic camera installed at the entrance of the limited space to obtain a first image;

[0049] An infrared camera installed in a limited space is used to collect an image in a low-light environment to obtain a second image;

[0050] A local close-up image is collected by using a high-resolution camera installed at an operating position of the equipment in a limited space to obtain a third image;

[0051] The second image and the third image are fused with the first image to obtain a panoramic image of the limited space operation.

[0052] In one embodiment, the training method of the deep learning model in the recognition result acquisition module includes:

[0053] Collect images of limited space work sites from different channels to obtain training image data;

[0054] Label the personal protective equipment and on-site safety facilities in the training image data to obtain a training data set; the personal protective equipment includes a safety helmet, a respirator and a safety belt, and the on-site safety facilities include fencing facilities, safety signs, warning signs, safety notice boards, forced air supply equipment, lighting equipment, intercom equipment, a tripod and a gas detection alarm.

[0055] The training data set is input into the deep learning model for training to obtain a trained deep learning model.

[0056] In one embodiment, inputting the training data set into the deep learning model for training to obtain a trained deep learning model includes:

[0057] The training data set is divided into training set, validation set and test set according to three ratios;

[0058] The training set is used to make the model learn the mapping relationship from input data to output data. During the learning process, the pre-trained weights on the training data set are used as the initial weights.

[0059] The performance of the model is evaluated using a validation set, and the hyperparameters are adjusted according to the performance evaluation results until the model performance reaches a predetermined effect, thereby obtaining an initial deep learning model; the hyperparameters include a learning rate and the number of iterations, and the learning rate gradually decreases during the training process;

[0060] The test set is used to evaluate the generalization performance of the initial deep learning model, and the initial deep learning model whose generalization performance meets the expected effect is selected as the trained deep learning model.

[0061] In one embodiment, the warning information generation module is specifically used to:

[0062] Identify equipment warning pools based on personal protective equipment and all equipment included in the on-site safety facility;

[0063] A corresponding warning switch is configured for each personal protective equipment and on-site safety facility; if the corresponding warning switch is in the on state, the corresponding personal protective equipment or on-site safety facility is added to the equipment warning pool; if the corresponding warning switch is in the off state, the corresponding personal protective equipment or on-site safety facility is removed from the equipment warning pool;

[0064] Assign an identification identifier to each confined space operation;

[0065] Setting the opening and closing of each warning switch according to the identification requirements of the confined space operation corresponding to each identity identifier;

[0066] Generate warning information on wearing of personal protective equipment according to the wearing status of personal protective equipment and the opening and closing status of the warning switch;

[0067] Generate early warning information on the setting of on-site safety facilities based on the setting of on-site safety facilities and the opening and closing status of early warning switches.

[0068] In one embodiment, the warning information generation module generates warning information about the wearing of personal protective equipment according to the wearing condition of the personal protective equipment and the on / off condition of the warning switch, including:

[0069] Generate a person target frame based on the identified persons in the limited space;

[0070] Identify personal protective equipment within the personnel target frame and obtain the wearing status of the personal protective equipment;

[0071] When the warning switch is set to the on state and the corresponding personal protective equipment is not recognized, warning information about the wearing of the personal protective equipment is generated.

[0072] In one embodiment, the warning information generation module generates warning information about the setting of the on-site safety facilities according to the setting of the on-site safety facilities and the opening and closing of the warning switch, including:

[0073] A flag bit is configured for each device included in the on-site safety facility. When the corresponding device is identified, the corresponding flag bit is set to 1; when the corresponding device is not identified, the corresponding flag bit is set to 0;

[0074] Initialize the flag bits of all devices to 0 at the beginning of limited space operation;

[0075] When the warning switch is set to the on state and the corresponding on-site safety facility flag is 0, a warning message about the on-site safety facility setting is generated;

[0076] When the warning switch is set to the on state and the corresponding on-site safety facility flag changes from 0 to 1, the warning information about the on-site safety facility setting generated before this operation is eliminated.

[0077] In a third aspect, the present invention provides an electronic device, comprising:

[0078] processor, memory, and an interface for communicating with a gateway;

[0079] The memory is used to store programs and data, and the processor calls the program stored in the memory to execute a method for identifying and warning safety hazards in confined space operations provided in any one of the first aspects.

[0080] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium includes a program, and when the program is executed by a processor, the program is used to execute a method for identifying and warning of safety hazards in confined space operations provided in any one of the first aspects.

[0081] The beneficial effects of the present invention are as follows: the present invention can fuse images captured by image acquisition devices dispersed at multiple locations of a limited space operation site in real time to generate a panoramic image, and automatically identify the configuration of personal protective equipment and safety facilities at the operation site through a deep learning model. Images from different angles and different locations are seamlessly spliced ​​through image fusion technology to form a complete panoramic image of limited space operations. The panoramic image is input into a pre-trained deep learning model, and the deep learning model can accurately identify personal protective equipment and on-site safety facilities in the image after a large amount of data training and optimization. The recognition result not only includes the presence or absence of these equipment and facilities, but also further determines the correctness of their wearing and setting. After obtaining the recognition result, the trigger conditions of the early warning signal are configured, and these conditions can be personalized according to different operating environments and safety requirements. The sending methods of the early warning signal are also diverse, and the early warning signal can be sent to relevant personnel in real time through multiple channels such as mobile devices, text messages, and emails to ensure that they can obtain the latest safety information in a timely manner. At the same time, the early warning signal will also be automatically transmitted to the control center so that the management personnel can fully understand the safety status of the operation site and take corresponding measures. The present invention greatly improves the safety and reliability of limited space operations. By real-time monitoring of the operation status, automatically identifying potential safety hazards and issuing early warnings, safety accidents caused by human factors can be effectively reduced. This not only protects the lives of operators, but also saves companies from economic losses caused by accidents. At the same time, it also provides a safer and more reliable working environment for confined space operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1The figure is a schematic flow chart of a method for identifying and warning safety hazards in confined space operations provided by an embodiment of the present invention;

[0083] Figure 2 The figure is a schematic structural diagram of a device for identifying and warning safety hazards in confined space operations provided by an embodiment of the present invention;

[0084] Figure 3 FIG. 1 is a schematic structural diagram of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0085] In order to make the purpose, technical solution and advantages of the present invention clearer and more understandable, the present invention is further described below in conjunction with the accompanying drawings and specific implementation methods. 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.

[0086] Based on the shortcomings of the prior art, the present invention provides a specific implementation method of a method for identifying and warning safety hazards in confined space operations, such as Figure 1 As shown, the method specifically includes:

[0087] S110: fusing images captured by image acquisition devices dispersed at multiple locations on the confined space operation site to obtain a panoramic image of the confined space operation.

[0088] Specifically, multiple image acquisition devices are reasonably arranged at the confined space operation site to ensure that they can cover every corner of the operation area. With the help of image fusion technology (such as weighted average method, principal component analysis method, multi-scale decomposition method, etc.), multiple image data from confined space operations are fused according to certain rules and strategies to obtain a panoramic image of the confined space operation through image fusion technology.

[0089] S120: Inputting the confined space operation panoramic image into a trained deep learning model to identify personal protective equipment and on-site safety facilities in the confined space operation panoramic image to obtain a recognition result; the recognition result includes the wearing status of the personal protective equipment and the setting status of the on-site safety facilities.

[0090] Specifically, the personal protective equipment and on-site safety facilities in the images are automatically identified by inputting panoramic images of confined space operations into a pre-trained deep learning model. First, an efficient and accurate deep learning model is built to identify personal protective equipment and on-site safety facilities; then, the model is trained with a large amount of annotated data, which includes panoramic images of various confined space operation scenarios, as well as the location and type of personal protective equipment and on-site safety facilities in these images; after the deep learning model is trained, new panoramic images of confined space operations can be input into the model to obtain recognition results.

[0091] S130: Configure trigger conditions for generating a warning signal, and generate a corresponding warning signal based on the recognition result.

[0092] Specifically, different trigger conditions are configured according to different confined space working environments to determine the equipment that needs to be monitored in personal protective equipment and safety facilities for confined space operations. If there is no corresponding equipment in the identification result, an early warning message that the corresponding equipment is not set is generated.

[0093] S140: Send the warning signal to the mobile device or control center through multiple channels.

[0094] Specifically, the warning signal can be quickly sent to mobile devices or monitoring centers through a variety of channels, such as sound, visual signals, and direct notification of responsible personnel. The target types and their bounding boxes detected by intelligent recognition technology can also be marked on the real-time video stream to provide immediate feedback. The warning information can also be transmitted to the on-site monitoring software in real time, and prompts can be given on-site through sound and light alarms.

[0095] In this embodiment, by real-time fusion of images captured by image acquisition devices at multiple locations to generate a panoramic image, the safety and reliability of the work site can be improved. The deep learning model is used to automatically identify the configuration of personal protective equipment and safety facilities at the work site to improve the recognition accuracy. By configuring the trigger conditions of the early warning signal, it can be personalized according to different working environments and safety requirements to improve the practicality of the early warning signal. In addition, the early warning signal is sent in a variety of ways to ensure that relevant personnel can obtain the latest safety information in a timely manner. The early warning signal can be automatically transmitted to the control center, which is convenient for managers to fully understand the safety status of the work site and take corresponding measures. It can realize real-time monitoring of the working status of the limited space, automatically identify safety hazards and issue early warnings, reduce safety accidents caused by human factors, protect the lives of workers, save economic losses for enterprises, and provide a safer and more reliable working environment for limited space operations.

[0096] In an embodiment of the present invention, S110 specifically includes the following steps:

[0097] The entire working area is captured by a panoramic camera installed at the entrance of the limited space to obtain a first image;

[0098] An infrared camera installed in a limited space is used to collect an image in a low-light environment to obtain a second image;

[0099] A local close-up image is collected by using a high-resolution camera installed at an operating position of the equipment in a limited space to obtain a third image;

[0100] The second image and the third image are fused with the first image to obtain a panoramic image of the limited space operation.

[0101] In this embodiment, a panoramic camera is installed at the entrance of the limited space. The panoramic camera has an ultra-wide-angle field of view and can capture the image of the entire working area at one time. The picture taken by the panoramic camera is clear and detailed, and the situation of the entire space can be grasped at all times. An infrared camera is installed inside the limited space. The infrared camera can form an image by capturing the infrared radiation emitted by the object in the case of insufficient light. It is particularly effective at night or in dark areas and can provide clear and reliable image data. In addition to the panoramic camera and the infrared camera, a high-resolution camera is also installed at the operating position of the limited space equipment. The high-resolution camera has extremely high pixels and resolution and can capture local close-up images of the equipment. Using image fusion technology, the second image and the third image taken by the infrared camera and the high-resolution camera are fused with the first image taken by the panoramic camera. Through precise image registration and fusion algorithms, the image information captured by different cameras is integrated to form a complete and comprehensive panoramic image of the limited space operation. This image not only contains the macro information of the entire working area, but also covers the local close-up of the equipment operating position and the image details in the low-light environment. It provides a comprehensive, detailed and real view of the working environment. In actual applications, it can provide operators with more comprehensive and detailed visual support, reduce operational difficulty and reduce human errors. By installing panoramic cameras, infrared cameras and high-resolution cameras and using image fusion technology, it can obtain comprehensive and detailed panoramic images of limited space operations, which not only improves the safety and accuracy of operations, but also brings more convenience and benefits to industrial production.

[0102] In one embodiment of the present invention, the training method of the deep learning model in S120 includes:

[0103] Collect images of limited space work sites from different channels to obtain training image data;

[0104] Label the personal protective equipment and on-site safety facilities in the training image data to obtain a training data set; the personal protective equipment includes but is not limited to safety helmets, respirators and safety belts, and the on-site safety facilities include but are not limited to fencing facilities, safety signs, warning signs, safety notice boards, forced air supply equipment, lighting equipment, intercom equipment, tripods and gas detection alarms.

[0105] The training data set is input into the deep learning model for training to obtain a trained deep learning model.

[0106] More specifically, the method of training a deep learning model using a training dataset is as follows:

[0107] The training data set is divided into training set, validation set and test set according to three ratios;

[0108] The training set is used to make the model learn the mapping relationship from input data to output data. During the learning process, the pre-trained weights on the training data set are used as the initial weights. The model can use the YOLOv8 model.

[0109] The performance of the model is evaluated using the validation set, and the hyperparameters are adjusted according to the performance evaluation results until the model performance reaches the expected effect, thereby obtaining the initial deep learning model; wherein the hyperparameters include the learning rate and the number of iterations, and the learning rate gradually decreases during the training process;

[0110] The test set is used to evaluate the generalization performance of the initial deep learning model, and the initial deep learning model whose generalization performance meets the expected effect is selected as the trained deep learning model.

[0111] In this embodiment, by marking multiple types of personal protective equipment and on-site safety facilities, the model can identify different types of personal protective equipment and on-site safety facilities. Using pre-trained weights as initial weights can significantly speed up the training speed of the model. By using pre-trained weights, the model can find the correct parameter direction faster, thereby accelerating the training process, greatly shortening the time of model training, and improving the efficiency of research or application. In addition, since the pre-trained weights are trained on a large amount of data, they already have a certain generalization ability. When these weights are used as initial weights, the model can more easily learn new features, thereby better adapting to the new data distribution and improving the generalization ability. During the model training process, the learning rate gradually decreases. As the learning rate decreases, the step size updated by the model in each iteration gradually decreases, which helps the model converge more stably in the later stage of training. The loss fluctuations during the training process are usually small, making it easier to converge to the optimal solution and improve the stability of the model. Gradually reducing the learning rate helps the model to better generalize to unseen data in the later stage of training and avoid overfitting. It can achieve better performance in a shorter time, accelerate the training process, and use a strategy of gradually reducing the learning rate to save computing resources and reduce training costs. The trained deep learning model can be used to identify personal protective equipment and on-site safety facilities and improve recognition accuracy.

[0112] In one embodiment of the present invention, S130 configures a trigger condition for generating a warning signal, and generating a corresponding warning signal based on the recognition result includes:

[0113] Identify equipment warning pools based on personal protective equipment and all equipment included in the on-site safety facility;

[0114] A corresponding early warning switch is configured for each personal protective equipment and on-site safety facility; if the corresponding early warning switch is in the on state, the corresponding personal protective equipment or on-site safety facility is added to the equipment early warning pool; if the corresponding early warning switch is in the off state, the corresponding personal protective equipment or on-site safety facility is removed from the equipment early warning pool; the early warning switches include: safety helmet early warning switch, respirator early warning switch, seat belt early warning switch, enclosure facility early warning switch, safety sign early warning switch, warning sign early warning switch, safety notice board early warning switch, forced air supply equipment early warning switch, lighting equipment early warning switch, intercom equipment early warning switch, tripod early warning switch and gas detection alarm early warning switch.

[0115] Assign an identification identifier to each confined space operation;

[0116] Setting the opening and closing of each warning switch according to the identification requirements of the confined space operation corresponding to each identity identifier;

[0117] Generate warning information on wearing of personal protective equipment according to the wearing status of personal protective equipment and the opening and closing status of the warning switch;

[0118] Generate early warning information on the setting of on-site safety facilities based on the setting of on-site safety facilities and the opening and closing status of early warning switches.

[0119] In this embodiment, the equipment warning pool is a virtual collection that includes all personal protective equipment and on-site safety facilities that are in an active monitoring state. A unique warning switch is configured for each personal protective equipment and on-site safety facility. The state of the warning switch (on or off) determines whether the equipment should be included in the warning pool to adapt to different confined space operation scenarios and enhance adaptability. At the same time, a unique identity identifier is assigned to each confined space operation. This identity identifier can not only help identify different operating areas, but also set the opening and closing of the warning switch according to the specific needs of the operation. When the wearing of personal protective equipment or the opening and closing state of the warning switch changes, it will immediately generate corresponding warning information. Not only does it improve the monitoring efficiency of personal protective equipment and on-site safety facilities, but it also greatly reduces the incidence of safety accidents.

[0120] In one embodiment of the present invention, generating warning information about the wearing of personal protective equipment according to the wearing condition of the personal protective equipment and the on / off condition of the warning switch includes:

[0121] Generate a person target frame based on the identified persons in the limited space;

[0122] Identify personal protective equipment within the personnel target frame and obtain the wearing status of the personal protective equipment;

[0123] When the warning switch is set to the on state and the corresponding personal protective equipment is not recognized, warning information about the wearing of the personal protective equipment is generated.

[0124] In this embodiment, the purpose is to achieve real-time monitoring and early warning of the wearing of personal protective equipment, which not only enhances the safety awareness of the staff, but also reduces the incidence of safety accidents. Accurately identify the personnel in the limited space and generate an independent target frame for each person. Next, the personal protective equipment is identified within the range of the personnel target frame, and the wearing of each person's personal protective equipment is recorded for subsequent analysis and early warning. When the early warning switch is set to the on state, the wearing of personal protective equipment is monitored in real time. Once it is found that the corresponding personal protective equipment in a certain person's target frame is not identified, an early warning message about the wearing of personal protective equipment is immediately generated. The early warning message will specify which person and which equipment is not worn. Taking the safety helmet as an example, if the safety helmet warning switch is turned on, and the safety helmet is not detected in the target range frame of the operator, an "operator does not wear the safety helmet correctly" warning is generated.

[0125] In one embodiment of the present invention, generating warning information of the on-site safety facility setting according to the setting of the on-site safety facility and the opening and closing of the warning switch includes:

[0126] A flag bit is configured for each device included in the on-site safety facility. When the corresponding device is identified, the corresponding flag bit is set to 1; when the corresponding device is not identified, the corresponding flag bit is set to 0;

[0127] Initialize the flag bits of all devices to 0 at the beginning of limited space operation;

[0128] When the warning switch is set to the on state and the corresponding on-site safety facility flag is 0, a warning message about the on-site safety facility setting is generated;

[0129] When the warning switch is set to the on state and the corresponding on-site safety facility flag changes from 0 to 1, the warning information about the on-site safety facility setting generated before this operation is eliminated.

[0130] In this embodiment, the purpose is to accurately generate corresponding warning information according to the specific configuration of the on-site safety facilities and the state of the warning switch, so as to take timely measures when necessary. First, a flag bit is configured for each device. When the corresponding device is successfully identified by the system, the flag bit will be automatically set to 1, indicating that the device is set normally; conversely, if the device is not identified, the flag bit will be set to 0, warning that there may be safety problems. Before each confined space operation begins, an initialization process is performed to set the flag bits of all devices to 0. The purpose of this is to ensure that the system can start from scratch and comprehensively and accurately monitor the settings of each device. Next, when the warning switch is set to the on state, the setting of the on-site safety facilities begins to be monitored in real time. If the flag bit of a device is found to be 0, that is, the device is not identified, a warning message about the setting of the on-site safety facilities will be immediately generated. The generation of warning information is not static. In the actual operation process, sometimes the device may not be identified for some reason, but then it returns to normal. In order to avoid interference caused by false alarms, when the warning switch is still in the on state and the flag of the device previously judged to be abnormal changes from 0 to 1, it is considered that the device has returned to normal, thereby eliminating the warning information about the device previously generated. Taking the enclosure facility as an example, if the enclosure facility is detected, the "enclosure facility" flag of the operation is set to 1; if the enclosure facility warning switch is turned on and the "enclosure facility" flag of the operation is 0, the warning "no enclosure facility is set at the operation site" is generated; if the enclosure facility warning switch is turned on and the flag is 1, the warning "no enclosure facility is set at the operation site" generated before this operation is eliminated. The mechanism of generating and eliminating warning information not only ensures the comprehensive monitoring of safety facilities and timely handling of potential hazards, but also reduces the negative impact caused by false alarms. By continuously optimizing and improving this mechanism, it is possible to further improve operating efficiency, reduce safety risks, and provide a safer and more reliable working environment for on-site staff.

[0131] Based on the same inventive concept, the embodiments of the present application also provide a device for identifying and warning safety hazards in confined space operations, which can be used to implement a method for identifying and warning safety hazards in confined space operations described in the above embodiments, as described in the following embodiments. Since the principle of solving the problem by a device for identifying and warning safety hazards in confined space operations is similar to that of a method for identifying and warning safety hazards in confined space operations, the implementation of a device for identifying and warning safety hazards in confined space operations can refer to the implementation of a method for identifying and warning safety hazards in confined space operations, and the repeated parts will not be repeated. As used below, the terms "unit" or "module" can be a combination of software and / or hardware that implements predetermined functions. Although the system described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0132] The present invention provides a device for identifying and warning safety hazards in confined space operations. Figure 2 As shown. Figure 2 The device comprises:

[0133] Panoramic image acquisition module 210: used to fuse images captured by image acquisition devices dispersed at multiple locations on the limited space operation site to obtain a panoramic image of the limited space operation;

[0134] Recognition result acquisition module 220: used to input the confined space operation panoramic image into the trained deep learning model to identify the personal protective equipment and on-site safety facilities in the confined space operation panoramic image, and obtain a recognition result; the recognition result includes the wearing condition of the personal protective equipment and the setting condition of the on-site safety facilities;

[0135] Warning information generation module 230: used to configure the triggering conditions for generating a warning signal, and to generate a corresponding warning signal based on the recognition result;

[0136] Warning information sending module 240: used to send warning signals to mobile devices or control centers through multiple channels.

[0137] In an embodiment of the present invention, the panoramic image acquisition module 210 is specifically used for:

[0138] The entire working area is captured by a panoramic camera installed at the entrance of the limited space to obtain a first image;

[0139] An infrared camera installed in a limited space is used to collect an image in a low-light environment to obtain a second image;

[0140] A local close-up image is collected by using a high-resolution camera installed at an operating position of the equipment in a limited space to obtain a third image;

[0141] The second image and the third image are fused with the first image to obtain a panoramic image of the limited space operation.

[0142] In one embodiment of the present invention, the training method of the deep learning model in the recognition result acquisition module 220 includes:

[0143] Collect images of limited space work sites from different channels to obtain training image data;

[0144] Label the personal protective equipment and on-site safety facilities in the training image data to obtain a training data set; the personal protective equipment includes a safety helmet, a respirator and a safety belt, and the on-site safety facilities include fencing facilities, safety signs, warning signs, safety notice boards, forced air supply equipment, lighting equipment, intercom equipment, a tripod and a gas detection alarm.

[0145] The training data set is input into the deep learning model for training to obtain a trained deep learning model.

[0146] In one embodiment of the present invention, inputting the training data set into the deep learning model for training to obtain a trained deep learning model includes:

[0147] The training data set is divided into training set, validation set and test set according to three ratios;

[0148] The training set is used to make the model learn the mapping relationship from input data to output data. During the learning process, the pre-trained weights on the training data set are used as the initial weights.

[0149] The performance of the model is evaluated using a validation set, and the hyperparameters are adjusted according to the performance evaluation results until the model performance reaches a predetermined effect, thereby obtaining an initial deep learning model; the hyperparameters include a learning rate and the number of iterations, and the learning rate gradually decreases during the training process;

[0150] The test set is used to evaluate the generalization performance of the initial deep learning model, and the initial deep learning model whose generalization performance meets the expected effect is selected as the trained deep learning model.

[0151] In one embodiment of the present invention, the warning information generation module 230 is specifically used to:

[0152] Identify equipment warning pools based on personal protective equipment and all equipment included in the on-site safety facility;

[0153] A corresponding warning switch is configured for each personal protective equipment and on-site safety facility; if the corresponding warning switch is in the on state, the corresponding personal protective equipment or on-site safety facility is added to the equipment warning pool; if the corresponding warning switch is in the off state, the corresponding personal protective equipment or on-site safety facility is removed from the equipment warning pool;

[0154] Assign an identification identifier to each confined space operation;

[0155] Setting the opening and closing of each warning switch according to the identification requirements of the confined space operation corresponding to each identity identifier;

[0156] Generate warning information on wearing of personal protective equipment according to the wearing status of personal protective equipment and the opening and closing status of the warning switch;

[0157] Generate early warning information on the setting of on-site safety facilities based on the setting of on-site safety facilities and the opening and closing status of early warning switches.

[0158] In one embodiment of the present invention, the warning information generation module 230 generates warning information about the wearing of personal protective equipment according to the wearing condition of the personal protective equipment and the on / off condition of the warning switch, including:

[0159] Generate a person target frame based on the identified persons in the limited space;

[0160] Identify personal protective equipment within the personnel target frame and obtain the wearing status of the personal protective equipment;

[0161] When the warning switch is set to the on state and the corresponding personal protective equipment is not recognized, warning information about the wearing of the personal protective equipment is generated.

[0162] In one embodiment of the present invention, the warning information generation module 230 generates warning information of the on-site safety facility settings according to the settings of the on-site safety facilities and the on-off status of the warning switch, including:

[0163] A flag bit is configured for each device included in the on-site safety facility. When the corresponding device is identified, the corresponding flag bit is set to 1; when the corresponding device is not identified, the corresponding flag bit is set to 0;

[0164] Initialize the flag bits of all devices to 0 at the beginning of limited space operation;

[0165] When the warning switch is set to the on state and the corresponding on-site safety facility flag is 0, a warning message about the on-site safety facility setting is generated;

[0166] When the warning switch is set to the on state and the corresponding on-site safety facility flag changes from 0 to 1, the warning information about the on-site safety facility setting generated before this operation is eliminated.

[0167] The embodiments of the present application also provide a specific implementation of an electronic device that can implement all steps in a method for identifying and warning safety hazards in confined space operations in the above embodiment, see Figure 3 , the electronic device 300 specifically includes the following contents:

[0168] Processor 310, memory 320, communication unit 330 and bus 340;

[0169] The processor 310 , the memory 320 , and the communication unit 330 communicate with each other via the bus 340 ; the communication unit 330 is used to implement information transmission between server-side devices and terminal devices and other related devices.

[0170] The processor 310 is used to call the computer program in the memory 320. When the processor executes the computer program, all the steps in the method for identifying and warning safety hazards in confined space operations in the above embodiment are implemented.

[0171] Those skilled in the art should understand that the memory may be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), etc. The memory is used to store programs, and the processor executes the programs after receiving the execution instruction. Furthermore, the software programs and modules in the above-mentioned memory may also include an operating system, which may include various software components and / or drivers for managing system tasks (such as memory management, storage device control, power management, etc.), and may communicate with various hardware or software components to provide an operating environment for other software components.

[0172] The processor may be an integrated circuit chip having the ability to process signals. The above-mentioned processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc. The methods, steps, and logic block diagrams disclosed in the embodiments of the present application may be implemented or executed. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0173] The present application also provides a computer-readable storage medium, which includes a program. When the program is executed by a processor, it is used to execute a method for identifying and warning of safety hazards in confined space operations provided by any of the aforementioned method embodiments.

[0174] Those skilled in the art should understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk, etc., various media that can store program codes, and the specific media type is not limited in this application.

[0175] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for identifying and warning safety hazards in confined space operations, characterized in that: The method comprises: The images captured by the image acquisition devices scattered at multiple locations on the limited space operation site are fused to obtain a panoramic image of the limited space operation; Inputting the confined space operation panoramic image into the trained deep learning model to identify the personal protective equipment and on-site safety facilities in the confined space operation panoramic image, and obtaining a recognition result; the recognition result includes the wearing condition of the personal protective equipment and the setting condition of the on-site safety facilities; Configure the trigger conditions for generating warning signals, and generate corresponding warning signals based on the recognition results; Send warning signals to mobile devices or control centers through multiple channels.

2. A method for identifying and warning safety hazards in confined space operations according to claim 1, characterized in that: The fusing of images captured by image acquisition devices dispersed at multiple locations on the limited space operation site to obtain a panoramic image of the limited space operation includes: The entire working area is captured by a panoramic camera installed at the entrance of the limited space to obtain a first image; An infrared camera installed in a limited space is used to collect an image in a low-light environment to obtain a second image; A local close-up image is collected by using a high-resolution camera installed at an operating position of the equipment in a limited space to obtain a third image; The second image and the third image are fused with the first image to obtain a panoramic image of the limited space operation.

3. A method for identifying and warning safety hazards in confined space operations according to claim 1, characterized in that: The training method of the deep learning model includes: Collect images of limited space work sites from different channels to obtain training image data; Label the personal protective equipment and on-site safety facilities in the training image data to obtain a training data set; the personal protective equipment includes a safety helmet, a respirator and a safety belt, and the on-site safety facilities include fencing facilities, safety signs, warning signs, safety notice boards, forced air supply equipment, lighting equipment, intercom equipment, a tripod and a gas detection alarm. The training data set is input into the deep learning model for training to obtain a trained deep learning model.

4. A method for identifying and warning safety hazards in confined space operations as claimed in claim 3, characterized in that: Inputting the training data set into the deep learning model for training to obtain a trained deep learning model includes: The training data set is divided into training set, validation set and test set according to three ratios; The training set is used to make the model learn the mapping relationship from input data to output data. During the learning process, the pre-trained weights on the training data set are used as the initial weights. The performance of the model is evaluated using a validation set, and the hyperparameters are adjusted according to the performance evaluation results until the model performance reaches a predetermined effect, thereby obtaining an initial deep learning model; the hyperparameters include a learning rate and the number of iterations, and the learning rate gradually decreases during the training process; The test set is used to evaluate the generalization performance of the initial deep learning model, and the initial deep learning model whose generalization performance meets the expected effect is selected as the trained deep learning model.

5. A method for identifying and warning safety hazards in confined space operations according to claim 1, characterized in that: The configuration of the triggering conditions for generating the warning signal, and generating the corresponding warning signal based on the recognition result, includes: Identify equipment warning pools based on personal protective equipment and all equipment included in the on-site safety facility; A corresponding warning switch is configured for each personal protective equipment and on-site safety facility; if the corresponding warning switch is in the on state, the corresponding personal protective equipment or on-site safety facility is added to the equipment warning pool; if the corresponding warning switch is in the off state, the corresponding personal protective equipment or on-site safety facility is removed from the equipment warning pool; Assign an identification identifier to each confined space operation; Setting the opening and closing of each warning switch according to the identification requirements of the confined space operation corresponding to each identity identifier; Generate warning information on wearing of personal protective equipment according to the wearing status of personal protective equipment and the opening and closing status of the warning switch; Generate early warning information on the setting of on-site safety facilities based on the setting of on-site safety facilities and the opening and closing status of early warning switches.

6. A method for identifying and warning safety hazards in confined space operations as claimed in claim 5, characterized in that: The generating of warning information on wearing of personal protective equipment according to the wearing condition of the personal protective equipment and the opening and closing condition of the warning switch comprises: Generate a person target frame based on the identified persons in the limited space; Identify personal protective equipment within the personnel target frame and obtain the wearing status of the personal protective equipment; When the warning switch is set to the on state and the corresponding personal protective equipment is not recognized, warning information about the wearing of the personal protective equipment is generated.

7. A method for identifying and warning safety hazards in confined space operations as claimed in claim 5, characterized in that: The generating of the warning information of the on-site safety facility setting according to the setting of the on-site safety facility and the opening and closing of the warning switch comprises: A flag bit is configured for each device included in the on-site safety facility. When the corresponding device is identified, the corresponding flag bit is set to 1; when the corresponding device is not identified, the corresponding flag bit is set to 0; Initialize the flag bits of all devices to 0 at the beginning of limited space operation; When the warning switch is set to the on state and the corresponding on-site safety facility flag is 0, a warning message about the on-site safety facility setting is generated; When the warning switch is set to the on state and the corresponding on-site safety facility flag changes from 0 to 1, the warning information about the on-site safety facility setting generated before this operation is eliminated.

8. A device for identifying and warning safety hazards in confined space operations, characterized in that: The device comprises: Panoramic image acquisition module: used to fuse images captured by image acquisition devices scattered at multiple locations on the limited space operation site to obtain a panoramic image of the limited space operation; Recognition result acquisition module: used to input the confined space operation panoramic image into the trained deep learning model to identify the personal protective equipment and on-site safety facilities in the confined space operation panoramic image, and obtain the recognition result; the recognition result includes the wearing condition of the personal protective equipment and the setting condition of the on-site safety facilities; Warning information generation module: used to configure the trigger conditions for generating warning signals and generate corresponding warning signals based on the recognition results; Warning information sending module: used to send warning signals to mobile devices or control centers through multiple channels.

9. An electronic device, characterized in that: include: processor, memory, and an interface for communicating with a gateway; The memory is used to store programs and data, and the processor calls the program stored in the memory to execute the method described in any one of claims 1 to 4.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a program, and when the program is executed by a processor, the program is used to perform the method according to any one of claims 1 to 4.

Images