A dangerous operation space safety production intelligent monitoring method and system

Abstract

This application provides an intelligent monitoring method for safe production in hazardous work spaces. The method includes: collecting multimodal image data and environmental parameter data of the hazardous work space; the multimodal image data includes infrared thermal imaging images, polarized light images, and visible light images; the environmental parameter data includes dust concentration, light intensity, smoke concentration, gas concentration, and temperature and humidity parameters; fusing the multimodal image data to obtain fused image data; performing target recognition processing based on the fused image data to obtain personnel behavior feature data and equipment operating status data; combining the personnel behavior feature data, equipment operating status data, and environmental parameter data to perform risk assessment calculation to obtain a comprehensive risk value; and triggering a corresponding early warning response operation based on the comprehensive risk value. This application solves the problem of difficult imaging in harsh environments by fusing infrared thermal imaging images, polarized light images, and visible light images.

Description

Technical Field

[0001] This application relates to the field of intelligent monitoring, specifically to an intelligent monitoring method and system for safe production in hazardous work spaces. Background Technology

[0002] Although the total number of work safety accidents has been declining in recent years, major and serious accidents still occur, highlighting structural risks. Safety risks are intertwined and compounded in key industries such as mining, hazardous chemicals, construction, and gas pipeline operations. Many work spaces present hazards such as flammable and explosive materials, toxic and harmful substances, and mechanical injuries. Traditional safety supervision relies on manual on-site inspections, characterized by sporadic and surprise inspections, making it difficult to achieve full-time, full-process monitoring coverage. Furthermore, some current monitoring solutions rely solely on visible light image acquisition for risk monitoring, with a core flaw being extremely poor environmental adaptability. Hazardous work spaces are often accompanied by harsh visual environments such as dust, smoke obstruction, and backlighting. Visible light images suffer from detail loss, blurred targets, and even imaging failure in these scenarios, failing to accurately identify non-compliant personnel behavior and abnormal equipment conditions. For example, the large amount of dust generated after underground blasting operations in mines can completely obscure visible light lenses, rendering the monitoring system ineffective and unable to effectively predict risks for subsequent operations. Even in normal working conditions, strong light reflection can easily misinterpret the metal surface of equipment as abnormal heating, leading to invalid risk warnings and disrupting normal regulatory processes. Furthermore, a single visual image can only capture surface features and cannot correlate with environmental parameters. Even if personnel are identified as performing standard operations, it is difficult to predict the potential risks of them being in a high-concentration gas environment, resulting in incomplete risk assessments. Summary of the Invention

[0003] To address the problem of one-sided risk assessment in the practical application of existing technologies, this application provides an intelligent monitoring method and system for safe production in hazardous work spaces.

[0004] The first aspect of this application provides an intelligent monitoring method for safe production in hazardous work spaces, comprising: Collect multimodal image data and environmental parameter data of hazardous work spaces. The multimodal image data includes infrared thermal imaging images, polarized light images, and visible light images. The environmental parameter data includes dust concentration, light intensity, smoke concentration, gas concentration, temperature and humidity parameters. The multimodal image data is fused to obtain fused image data; Based on the fused image data, target recognition processing is performed to obtain personnel behavior feature data and equipment operating status data; By combining the aforementioned personnel behavior data, equipment operating status data, and environmental parameter data, a risk assessment calculation is performed to obtain a comprehensive risk value; The corresponding early warning response operation is triggered based on the comprehensive risk value.

[0005] In a possible implementation, the fusion processing of the multimodal image data to obtain fused image data includes: The infrared thermal imaging image, polarized light image, and visible light image are respectively normalized and Gaussian filtered to obtain preprocessed image data for each modality. The corresponding features were extracted from the preprocessed modal image data to obtain temperature difference features, material contour features and color texture features, respectively. The environmental complexity is calculated based on the environmental parameter data, and the weights of each modal feature are assigned according to the environmental complexity; the weights of each modal feature include: infrared feature weight, polarization feature weight, and visible light feature weight; The temperature difference features, material contour features, and color texture features are fused and reconstructed using the modal feature weights to obtain the fused image data.

[0006] In a possible implementation, calculating the environmental complexity based on the environmental parameter data and assigning weights to each modal feature according to the environmental complexity includes: The dust concentration, light intensity, and smoke concentration were normalized to obtain normalized dust concentration, normalized light intensity, and normalized smoke concentration, respectively. The environmental complexity is obtained by multiplying the normalized dust concentration, normalized light intensity, and normalized smoke concentration by their respective preset weighting coefficients and then summing the results.

[0007] In a possible implementation, calculating the environmental complexity based on the environmental parameter data and assigning weights to each modal feature according to the environmental complexity includes: The infrared feature weights and polarization feature weights are determined based on preset base values ​​and the environmental complexity. The weights of visible light features are determined based on environmental complexity.

[0008] In a possible implementation, the step of fusing and reconstructing the temperature difference features, material contour features, and color texture features according to assigned weights to obtain the fused image data includes: The temperature difference features, material contour features, and color texture features are multiplied by the corresponding modal feature weights, and then the product results are superimposed to obtain the pixel values ​​of each pixel in the fused image data, thus forming complete fused image data.

[0009] In a possible implementation, the target recognition processing based on the fused image data to obtain personnel behavior feature data and equipment operating status data includes: The environmental complexity calculated based on the environmental parameter data generates the recognition confidence threshold and the anchor frame size of the target recognition model; The target recognition model is adjusted based on the anchor frame size and the confidence threshold to generate a first recognition model; The first target recognition model is used to identify the fused image data to obtain the personnel behavior feature data and equipment operation status data.

[0010] In a possible implementation, the risk assessment calculation, which combines the personnel behavior characteristic data, equipment operating status data, and environmental parameter data to obtain a comprehensive risk value, includes: Construct a set of personnel behavioral characteristics, which includes four parameters: personnel protective equipment wearing status, personnel entry status into risk areas, normalized result of personnel stay time, and personnel operation standardization. An environmental parameter set is constructed, which includes four parameters: normalized gas concentration, normalized temperature and humidity deviation, normalized dust concentration, and normalized light intensity. Assign corresponding weight coefficients to each parameter in the set of personnel behavioral characteristics and each parameter in the set of environmental parameters. Multiply the inverse values ​​of each parameter of personnel behavioral characteristics by their corresponding weight coefficients and sum them. Then multiply each parameter of environmental parameters by their corresponding weight coefficients and sum them. Add the two sums together to obtain the comprehensive risk value.

[0011] In a possible implementation, the step involves assigning corresponding weight coefficients to each parameter in the personnel behavior characteristic set and each parameter in the environmental parameter set, multiplying the inverse values ​​of each personnel behavior characteristic parameter by their corresponding weight coefficients and summing the results, then multiplying each environmental parameter by its corresponding weight coefficient and summing the results, and finally adding the two sums to obtain the comprehensive risk value, including: The parameters in the set of personnel behavior characteristics are converted into inverse values. The converted inverse values ​​of personnel behavior characteristics are multiplied by the corresponding weight coefficients to obtain the risk contribution value of each behavior. The summation is then used to obtain the total behavior risk value. Multiply each parameter in the set of environmental parameters by its corresponding weight coefficient to obtain the contribution value of each environmental risk, and sum them to obtain the total environmental risk value; The total behavioral risk value is added to the total environmental risk value to obtain the comprehensive risk value.

[0012] In a possible implementation, triggering the corresponding early warning response operation based on the comprehensive risk value to complete the intelligent monitoring of safe production in hazardous work spaces includes: If the comprehensive risk value is greater than or equal to the first preset risk threshold, a level one warning is triggered, and on-site voice risk prompts, personnel location push notifications, and emergency department linkage operations are simultaneously initiated. If the comprehensive risk value is greater than or equal to the second preset risk threshold and less than the first preset risk threshold, a level two warning is triggered, and a warning message is pushed to the terminal device of the person in charge at the site, triggering a verification instruction. If the overall risk value is less than the second preset risk threshold, maintain normal monitoring status and transmit all data within this monitoring cycle to the storage module.

[0013] The second aspect of this application provides an intelligent monitoring system for safe production in hazardous work spaces, comprising: Multimodal data acquisition module: used to acquire multimodal image data and environmental parameter data of hazardous work spaces. The multimodal image data includes infrared thermal imaging images, polarized light images and visible light images. The environmental parameter data includes dust concentration, light intensity, smoke concentration, gas concentration, temperature and humidity parameters. Image fusion processing module: used to fuse the multimodal image data to obtain fused image data; The target recognition module is used to perform target recognition processing based on the fused image data to obtain personnel behavior feature data and equipment operating status data; The risk assessment module is used to combine the personnel behavior characteristic data, equipment operating status data, and environmental parameter data to perform risk assessment calculations and obtain a comprehensive risk value. The early warning response module is used to trigger corresponding early warning response operations based on the comprehensive risk value, thereby completing intelligent monitoring of safe production in hazardous work spaces.

[0014] The infrared thermal imaging, polarized light, and visible light images acquired in this application can penetrate obstructions to capture temperature differences, distinguish materials and contours against strong light, and reveal color details. Image fusion using these infrared thermal imaging, polarized light, and visible light images solves the problem of difficult imaging in harsh environments. Furthermore, this application uses parameters such as dust concentration and light intensity collected by sensors to directly reflect the environmental safety status, providing a basis for image fusion weight allocation and environmental parameters. Subsequently, this application performs target recognition processing on the fused image, extracting personnel behavior feature data and equipment operating status data. Based on the generated feature data and environmental parameter data, risks in the environment can be assessed, thereby triggering corresponding early warning measures. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating an intelligent monitoring method for safe production in hazardous work spaces, as described in an embodiment of this application.

[0017] Figure 2 This is a schematic diagram of the structure of an intelligent monitoring system for safe production in hazardous work spaces, as described in an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] It should be noted that underground mines and gas pipeline work areas typically present complex environmental parameters such as dust, smoke, poor visual conditions due to backlighting, toxic gases, and fluctuating temperature and humidity, as well as strong coupling between personnel and equipment risks. Existing monitoring technologies for hazardous work areas have the following shortcomings: 1. They rely solely on visible light video monitoring, resulting in insufficient accuracy in adverse visual environments. 2. Personnel behavior, equipment status, and environmental parameters are independent, making multi-dimensional risk assessment impossible. 3. They only issue risk warnings after a risk occurs, lacking pre-emptive capabilities. Therefore, there is an urgent need for an intelligent monitoring solution that can adapt to harsh environments and achieve multi-information fusion and risk prediction.

[0020] Based on this, such as Figure 1 As shown, this application provides an embodiment of an intelligent monitoring method for safe production in hazardous work spaces, the method comprising: S101, Collect multimodal image data and environmental parameter data of the hazardous work space. The multimodal image data includes infrared thermal imaging images, polarized light images and visible light images. The environmental parameter data includes dust concentration, light intensity, smoke concentration, gas concentration, temperature and humidity parameters.

[0021] S102, the multimodal image data is fused by combining the environmental parameter data to obtain fused image data.

[0022] S103, target recognition processing is performed based on the fused image data to obtain personnel behavior feature data and equipment operating status data.

[0023] S104. Combining the personnel behavior characteristic data, equipment operation status data, and environmental parameter data, a risk assessment calculation is performed to obtain a comprehensive risk value.

[0024] S105, trigger the corresponding early warning response operation based on the comprehensive risk value.

[0025] It should be noted that multimodal image data includes infrared thermal images, polarized light images, and visible light images. Infrared thermal images are acquired using infrared detectors. Based on the principle that all objects radiate infrared radiation, and that radiation intensity is positively correlated with temperature, they can capture differences in the thermal radiation of objects and penetrate obstructions such as dust and smoke. Polarized light images are acquired using polarization cameras. Based on the principle that different objects reflect or scatter light in different polarization states, they can distinguish the material and outline of objects and suppress strong light reflection. Visible light images are acquired using ordinary optical cameras. Based on the principle of direct light imaging, their function is to reveal the color and detailed texture of objects.

[0026] It should be noted that the environmental parameter data includes dust concentration, light intensity, smoke concentration, gas concentration, temperature and humidity parameters. The environmental parameter data is collected by corresponding sensors, such as dust sensors and gas detectors, and reflects the environmental status of the work space, providing a basis for subsequent environmental complexity calculation and risk assessment.

[0027] The infrared thermal imaging, polarized light, and visible light images acquired in this application can penetrate obstructions to capture temperature differences, distinguish materials and contours against strong light, and reveal color details. Image fusion using these infrared thermal imaging, polarized light, and visible light images solves the problem of difficult imaging in harsh environments. Furthermore, this application uses parameters such as dust concentration and light intensity collected by sensors to directly reflect the environmental safety status, providing a basis for image fusion weight allocation and environmental parameters. Subsequently, this application performs target recognition processing on the fused image, extracting personnel behavior feature data and equipment operating status data. Based on the generated feature data and environmental parameter data, risks in the environment can be assessed, thereby triggering corresponding early warning measures.

[0028] In one embodiment of this application, the step of fusing the multimodal image data with the environmental parameter data to obtain fused image data includes: S201, normalize and Gaussian filter the infrared thermal imaging image, polarized light image and visible light image respectively to obtain preprocessed image data of each modality. S202, extract corresponding features from the preprocessed modal image data to obtain temperature difference features, material contour features and color texture features respectively; S203, calculate the environmental complexity based on the environmental parameter data, and assign weights to each modal feature according to the environmental complexity; the weights of each modal feature include: infrared feature weight, polarization feature weight, and visible light feature weight; S204, the temperature difference features, material contour features and color texture features are fused and reconstructed using the modal feature weights to obtain the fused image data.

[0029] It should be noted that the grayscale ranges of images from different modalities vary significantly. Direct fusion can lead to an imbalance in feature proportions. Normalization processing involves adjusting the pixel grayscale values ​​of each image. Adjust to the range of 0-1; the specific normalization calculation formula is as follows: ,in, The minimum grayscale value of the image. This represents the maximum grayscale value of the image.

[0030] It should be noted that noise interference is likely to occur during image acquisition. In order to eliminate noise while preserving core features, this application uses a weighted calculation process with Gaussian filtering to eliminate noise. The specific calculation formula is as follows: ,in, Let be the Gaussian kernel function, where For the window radius, Standard deviation, Each pixel within the window is relative to the center pixel of the window. coordinate offset, Each pixel within the window is relative to the center pixel of the window. The coordinate offset, the window radius, and the standard deviation are adjusted according to the scenario, and this application does not impose any restrictions on them.

[0031] It should be noted that the temperature difference feature is extracted from the infrared image, reflecting the difference in thermal radiation of the object, and is used to highlight high-temperature targets. The material contour feature is extracted from the polarization image, fusing the degree of polarization and the polarization angle, and is used to distinguish and identify the material and contour of the object. The color texture feature is extracted from the visible light image, and is used to present color and detailed texture.

[0032] Compared to single-modal images, fused images can still clearly present target information in harsh environments such as dust, smoke, and backlight, such as whether personnel are wearing protective equipment or whether equipment is overheating abnormally. Compared to simple stitching fusion, this step in this application avoids feature redundancy and conflict, resulting in more accurate image information and improving the accuracy of subsequent target recognition.

[0033] In one embodiment of this application, the step of calculating environmental complexity based on the environmental parameter data and allocating weights for each modal feature according to the environmental complexity includes: S301, the dust concentration, light intensity, and smoke concentration are normalized respectively to obtain normalized dust concentration, normalized light intensity, and normalized smoke concentration. S302, the normalized dust concentration, normalized light intensity, and normalized smoke concentration are multiplied by their respective preset weighting coefficients, and the results of each calculation are added together to obtain the environmental complexity.

[0034] It should be noted that environmental complexity It is an indicator that quantifies the severity of visual pollution, and the specific formula is as follows: ,in, , , These are the normalized parameters for dust, light, and smoke. , , The weights are preset. The visual severity varies in different environments. Fixed weights have poor adaptability. Presetting different weight values ​​can more accurately calculate the complexity of the environment.

[0035] In one embodiment of this application, the step of calculating environmental complexity based on the environmental parameter data and allocating weights for each modal feature according to the environmental complexity includes: S401, determine the infrared feature weight and polarization feature weight based on the preset base value and the environmental complexity; S402, determine the visible light feature weights based on environmental complexity.

[0036] It should be noted that when assigning weights to each modal feature based on environmental complexity, the adaptability of different modal images in different environments must be considered. Specifically, the infrared feature weights are adjusted positively according to environmental complexity; the higher the environmental complexity, the greater the infrared feature weights. The calculation formula is as follows: Here, 0.3 is the base value for the infrared feature, ensuring that the infrared feature can play a fundamental role even in optimal environmental conditions. The polarization feature weight is negatively correlated with environmental complexity; the lower the environmental complexity, the larger the polarization feature weight. The calculation formula is: Where 0.3 is the base value for polarization characteristics. The visible light feature weight is also negatively correlated with environmental complexity, and the calculation formula is: This dynamic weight allocation ensures that each modality image can play its full role in its corresponding advantageous environment.

[0037] In one embodiment of this application, the step of fusing and reconstructing the temperature difference features, material contour features, and color texture features according to assigned weights to obtain the fused image data includes: The temperature difference features, material contour features, and color texture features are multiplied by the corresponding modal feature weights, and then the product results are superimposed to obtain the pixel values ​​of each pixel in the fused image data, thus forming complete fused image data.

[0038] In this embodiment, image fusion reconstruction is a process of integrating various modal features according to their weights. Specifically, temperature difference features, material contour features, and color texture features are multiplied by their corresponding modal feature weights, and then the product results are summed to obtain the pixel value of each pixel in the fused image data. The specific calculation formula is as follows: ,in To merge images at pixel points Pixel value at that location, Temperature difference features at pixels eigenvalues ​​at , for material contour features at pixels eigenvalues ​​at that location Color texture features at pixels The feature values ​​at the location can be used to form complete fused image data that simultaneously includes temperature, material, and color information.

[0039] In one embodiment of this application, target recognition processing is performed based on the fused image data to obtain personnel behavior feature data and equipment operating status data, including: S501, Based on the environmental parameter data, the environmental complexity is calculated to generate the recognition confidence threshold and the anchor frame size of the target recognition model; S502, the target recognition model is adjusted based on the anchor frame size and the confidence threshold to generate a first recognition model; S503, the first target recognition model is used to identify the fused image data to obtain the personnel behavior feature data and equipment operation status data.

[0040] It should be noted that the recognition confidence threshold is a key parameter for determining the validity of a target. It is dynamically adjusted based on environmental parameter data. The harsher the environment, the more blurred the target features in the fused image, making misidentification more likely. Therefore, the higher the environmental complexity, the higher the recognition confidence threshold. The confidence threshold is calculated as follows: The threshold is 0.5, which is the base threshold and 0.2 is the preset adjustment coefficient. Increasing the threshold can reduce misidentification caused by fuzzy features and ensure the reliability of the recognition results.

[0041] It should be noted that in this application, the anchor frame size is used to match the actual size of the target in the image and is adjusted based on environmental complexity. The higher the environmental complexity, the more easily the target outline in the fused image is deformed or blurred, and the anchor frame size needs to be appropriately increased to ensure that the target can be effectively selected. The formula for calculating the anchor frame size is as follows: ,in The base anchor frame size is 0.3, which is a preset adjustment coefficient. This adjustment can improve the adaptability of the target recognition model to deformable targets in harsh environments.

[0042] It should be noted that the target recognition model, such as the improved YOLOv8 model, with the adjusted recognition confidence threshold and anchor frame size is used to recognize the fused image data. During the recognition process, the model accurately detects people and equipment in the image based on the optimized parameters, extracts personnel behavior feature data such as the wearing status of protective equipment and the entry status of risk areas, as well as equipment operating status data such as whether the equipment is abnormally overheating and whether it is in a normal operating position, to provide structured data support for subsequent risk assessment.

[0043] The dynamic parameters generated by S501 in this application can automatically adjust the target recognition model and generate a first recognition model that adapts to the current environment. Compared with the fixed parameter mode of traditional models, the adjustment mechanism of this application can keep the model in the optimal recognition state at all times. For example, when the dust in the mine increases suddenly, the model can automatically increase the anchor frame and increase the threshold according to the environmental parameters. After the environment recovers, the parameters are automatically adjusted back to avoid the model's recognition performance from declining due to environmental changes, thus improving the robustness of the model in complex environments.

[0044] In one embodiment of this application, the step of combining the personnel behavior characteristic data, equipment operating status data, and environmental parameter data to perform risk assessment calculations and obtain a comprehensive risk value includes: S601, Construct a set of personnel behavior characteristics, which includes four parameters: personnel protective equipment wearing status, personnel risk area entry status, personnel stay time normalization result, and personnel operation standardization degree. S602, Construct an environmental parameter set, which includes four parameters: gas concentration normalization result, temperature and humidity deviation normalization result, dust concentration normalization result, and light intensity normalization result. S603 assigns corresponding weight coefficients to each parameter in the personnel behavior characteristic set and each parameter in the environmental parameter set. The reverse values ​​of each parameter of the personnel behavior characteristics are multiplied by their corresponding weight coefficients and then summed. Then, each parameter of the environmental parameters is multiplied by its corresponding weight coefficient and then summed. The two sums are added together to obtain the comprehensive risk value.

[0045] It should be noted that the personnel behavior characteristic set is a comprehensive quantification of personnel's operational safety status, including four parameters: personnel's protective equipment wearing status, personnel's entry status into risk areas, normalized results of personnel's stay time, and personnel's operational standardization. All of these parameters are derived from target identification results. Specifically, personnel's protective equipment wearing status is determined by identifying whether personnel are wearing safety helmets, protective clothing, etc.; personnel's entry status into risk areas is determined by comparing personnel's location with the preset risk area range; furthermore, personnel's stay time is obtained by continuously monitoring and normalizing the duration of their stay in specific areas; and personnel's operational standardization is determined by comparing their actions with standard operating procedures. These four parameters together constitute a complete assessment dimension for personnel behavior safety.

[0046] It should be noted that the environmental parameter set is used to quantify the environmental safety status of the workspace, including four parameters: normalized gas concentration, normalized temperature and humidity deviation, normalized dust concentration, and normalized light intensity. These parameters are all obtained by processing raw environmental data collected by sensors, providing core environmental data for risk assessment.

[0047] It should be noted that the comprehensive risk value integrates multi-dimensional risks through weighted summation. This involves assigning corresponding weight coefficients to each parameter in the set of personnel behavioral characteristics and each parameter in the set of environmental parameters. The inverse values ​​of each personnel behavioral characteristic parameter are multiplied by their corresponding weight coefficients and then summed to obtain the total behavioral risk value. The safer the state represented by a parameter, the smaller the inverse value; the more dangerous the state represented by a parameter, the larger the inverse value. Next, each environmental parameter is multiplied by its corresponding weight coefficient and then summed to obtain the total environmental risk value. Finally, the total behavioral risk value and the total environmental risk value are added together to obtain the comprehensive risk value. The specific calculation formula is as follows: ,in, For the comprehensive risk value, These are the behavioral feature weighting coefficients. For personnel behavior characteristic parameters, These are the environmental parameter weighting coefficients. These are environmental parameters.

[0048] For example, the weighting of personnel behavior characteristic parameters should highlight key safety factors. The weighting coefficients for personnel protective equipment wearing status and personnel entering risk areas are both 0.3, the weighting coefficient for the normalized result of personnel stay time is 0.2, and the weighting coefficient for personnel operation standardization is 0.2. In the weighting of environmental parameters, in order to highlight the core impact of toxic and harmful gases on safety, the weighting coefficient for the normalized result of gas concentration is 0.4, and the weighting coefficients for the normalized result of temperature and humidity deviation, the normalized result of dust concentration, and the normalized result of light intensity are all 0.2. Scientific weighting ensures the accuracy of risk assessment.

[0049] In one embodiment of this application, the steps involve configuring corresponding weight coefficients for each parameter in the personnel behavior feature set and each parameter in the environmental parameter set, multiplying the inverse values ​​of each personnel behavior feature parameter by their corresponding weight coefficients and summing the results, then multiplying each environmental parameter by its corresponding weight coefficient and summing the results, and finally adding the two sums to obtain the comprehensive risk value. S701, convert each parameter in the personnel behavior feature set into inverse values, multiply the converted personnel behavior feature inverse values ​​by the corresponding weight coefficients to obtain the risk contribution value of each behavior, and sum them to obtain the total behavior risk value; S702, multiply each parameter in the environmental parameter set by its corresponding weight coefficient to obtain the environmental risk contribution value of each parameter, and sum them to obtain the total environmental risk value; S703, the total behavioral risk value is added to the total environmental risk value to obtain the comprehensive risk value.

[0050] It should be noted that, since the original representation logic of personnel behavior characteristic parameters is opposite to the risk contribution logic, in the original characteristic parameters, the larger the value, the safer the personnel behavior. For example, suppose the corresponding behavior characteristic parameter value for wearing protective equipment is 1, and the corresponding behavior characteristic parameter value for not wearing protective equipment is 0. However, the inventors of this application have found that risk calculation should reflect that the more dangerous the behavior, the smaller the inverse value. Therefore, this application converts each parameter in the personnel behavior characteristic set into an inverse value. The specific conversion rule is: the safer the state represented by the parameter, the smaller the inverse value; the more dangerous the state represented by the parameter, the larger the inverse value. For example, if the corresponding behavior characteristic parameter for wearing protective equipment is 1, and the corresponding behavior characteristic parameter for not wearing protective equipment is 0, the converted inverse value is 1 for not wearing protective equipment and 0 for wearing protective equipment, ensuring that the risk contribution value is positively correlated with the degree of danger. Subsequently, the converted inverse values ​​of personnel behavior characteristics are multiplied by the corresponding weight coefficients. Weighting coefficients reflect the degree of impact of different behavioral parameters on overall safety risk. For example, the weight of protective equipment wearing status is higher than that of operational standardization, as its impact on personnel safety is more direct. Multiplication yields the risk contribution value of each individual behavior, quantifying the risk of a single behavior. Finally, the total behavioral risk value is obtained by summing all behavioral risk contribution values. This value comprehensively reflects the overall safety risk level brought about by the personnel's current work behavior, providing core data support for subsequent comprehensive risk assessments from a behavioral perspective.

[0051] In one embodiment achievable under this application, triggering a corresponding early warning response operation based on the comprehensive risk value to complete intelligent monitoring of safe production in hazardous work spaces includes: S801, if the comprehensive risk value is greater than or equal to the first preset risk threshold, a level one warning is triggered, and on-site voice risk prompts, personnel location push and emergency department linkage operations are started simultaneously. S802, if the comprehensive risk value is greater than or equal to the second preset risk threshold and less than the first preset risk threshold, a level two warning is triggered, and the warning information is pushed to the on-site person in charge's terminal device to trigger a verification instruction; S803: If the overall risk value is less than the second preset risk threshold, maintain normal monitoring status and transmit all data within the current monitoring cycle to the storage module.

[0052] For example, when the comprehensive risk value When the overall risk value is high, it indicates an extremely high safety risk in the work area, with the possibility of an accident at any time, requiring a Level 1 warning. At this point, a high-decibel voice risk alert is activated, prompting workers to evacuate immediately. The personnel's location is pushed to the relevant personnel's terminals to ensure accurate positioning of rescue personnel, coordinate with emergency departments, and activate the emergency response plan to minimize accident losses. When the overall risk value is high, it indicates a high safety risk in the work area, requiring immediate verification and handling. This triggers a Level 2 warning, pushing the warning information to the on-site supervisor's terminal device, specifying the warning location, risk type, and relevant parameters, and triggering a verification instruction. The on-site supervisor is required to go to the site within a specified time to verify the hazard and implement rectification to prevent further escalation of the risk. When the time is right, it indicates that the workspace is in a safe state and normal monitoring is maintained. All data within the current monitoring cycle is automatically recorded to the storage module, forming a complete monitoring archive. This facilitates subsequent traceability, verification, and data statistical analysis, providing data support for optimizing monitoring strategies.

[0053] In a second aspect of this application, an intelligent monitoring system for safe production in hazardous work spaces is proposed, the system comprising: Multimodal data acquisition module: used to acquire multimodal image data and environmental parameter data of hazardous work spaces. The multimodal image data includes infrared thermal imaging images, polarized light images and visible light images. The environmental parameter data includes dust concentration, light intensity, smoke concentration, gas concentration, temperature and humidity parameters. Image fusion processing module: used to fuse the multimodal image data to obtain fused image data; The target recognition module is used to perform target recognition processing based on the fused image data to obtain personnel behavior feature data and equipment operating status data; The risk assessment module is used to combine the personnel behavior characteristic data, equipment operating status data, and environmental parameter data to perform risk assessment calculations and obtain a comprehensive risk value. The early warning response module is used to trigger corresponding early warning response operations based on the comprehensive risk value, thereby completing intelligent monitoring of safe production in hazardous work spaces.

[0054] In one embodiment of this application, the intelligent monitoring system further includes seven modules: a basic data management module, a scene configuration module, a video monitoring module, a data dashboard module, an early warning management module, a hierarchical control module, and a situation analysis module, wherein: The basic data management module is a maintenance and management module for basic data such as operators, monitoring equipment, hazard sources, hazard points, and safety monitoring algorithms. It pre-enters and archives relevant elements of the work area to facilitate the allocation of elements and the scheduling of resources for handling hazards during subsequent safety supervision.

[0055] The scenario configuration module is the process of setting up regulatory personnel, regulatory processes, and regulatory content before the start of supervision. Different regulatory scenarios are configured according to the actual production environment and operation content.

[0056] The video monitoring module provides remote online video of the work site, allowing dispatch center managers to intuitively grasp the situation and monitor the work process at any time. The data dashboard module provides a comprehensive, multi-view, and dynamic display of safety production elements during the work process, enabling users to view the safety production status and development trends within the monitored area from multiple perspectives.

[0057] The early warning management module is an early warning mechanism for risk alerts generated during the operation supervision process. It sends basic emergency response information, such as risk alerts, risk levels, and relevant contingency plans, to the person in charge of production safety in the area through means such as lights, voice, and text messages.

[0058] The hierarchical management module includes six layers for handling safety hazards that arise during operations. Different handling procedures are formulated according to different safety hazard warning levels, linking relevant departments and personnel to respond in a timely manner and eliminate the probability of safety accidents.

[0059] The situation analysis module displays the overall safety production elements, including dynamic displays of monitoring indicators for various safety production elements and big data predictions of the trends and types of safety production hazards.

[0060] Traditional regulatory models rely on manual inspections and post-event handling, resulting in delayed responses and limited coverage. This application utilizes technologies such as the Internet of Things, artificial intelligence, and laser gas detection to achieve real-time data collection and analysis. Regulatory personnel can remotely monitor key indicators, and the system automatically identifies abnormal data and triggers risk alerts, reducing manual inspection time. This application not only speeds up problem handling but also reduces labor costs, allowing regulatory resources to be more concentrated in high-risk areas.

[0061] In addition, this application also realizes multi-level control functions for safety production supervision and emergency response, solves the problem of data silos in traditional supervision, and enables departments such as safety production, market supervision, and environmental supervision to share data and respond jointly in real time, breaking down data barriers between different departments, forming a joint regulatory force, and greatly improving the overall emergency response capability.

[0062] In another possible embodiment of this application, such as Figure 2 As illustrated, an exemplary intelligent monitoring system according to this application includes: an IoT sensing layer, a data analysis layer, an intelligent decision-making layer, and an application display layer. The IoT sensing layer collects raw data (video, personnel status, environmental parameters, etc.) from the site using devices such as mobile video sensors, smart safety helmets, wearable gas detectors, and temperature and humidity sensors. The collected data is then aggregated at a safety production terminal and transmitted to the upper-level data analysis layer via data aggregation. The data analysis layer performs basic processing on the audio and video data collected by the IoT layer using AI video analysis, voice parsing, and video decoding technologies. It uses distributed databases (HBase / Redis, Solr / ES) to support the storage and retrieval of various types of data. Furthermore, the data analysis layer provides map data, algorithm support, and data caching. After receiving and processing / storing the raw data from the IoT sensing layer, the data analysis layer supports the functions of the intelligent decision-making layer. The intelligent decision-making layer enables functions such as stream computing services, knowledge graphs, structured data, emergency plans, video surveillance, behavioral analysis, data perception, and accident early warning. Based on the processing results of the data analysis layer, it performs business logic processing and risk assessment, and then transmits the risk assessment results to the application presentation layer. The application presentation layer achieves visualized supervision, hazard early warning, hierarchical control, and situational analysis through video surveillance, multi-level early warning, hidden danger early warning, and monitoring of major hazard sources. It presents the processing results received from the intelligent decision-making layer to the user in a visual manner, supporting safety management decisions.

[0063] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for intelligent monitoring of safe production in hazardous work spaces, characterized in that, The method includes: Collect multimodal image data and environmental parameter data of hazardous work spaces. The multimodal image data includes infrared thermal imaging images, polarized light images, and visible light images. The environmental parameter data includes dust concentration, light intensity, smoke concentration, gas concentration, temperature and humidity parameters. The multimodal image data is fused by combining the environmental parameter data to obtain fused image data; Based on the fused image data, target recognition processing is performed to obtain personnel behavior feature data and equipment operating status data; By combining the aforementioned personnel behavior data, equipment operating status data, and environmental parameter data, a risk assessment calculation is performed to obtain a comprehensive risk value; The corresponding early warning response operation is triggered based on the comprehensive risk value.

2. The intelligent monitoring method for safe production in hazardous work spaces according to claim 1, characterized in that, The process of fusing the multimodal image data with the environmental parameter data to obtain fused image data includes: The infrared thermal imaging image, polarized light image, and visible light image are respectively normalized and Gaussian filtered to obtain preprocessed image data for each modality. The corresponding features are extracted from the preprocessed image data of each modality to generate each modal feature, which includes temperature difference features, material contour features and color texture features; The environmental complexity is calculated based on the environmental parameter data, and the weights of each modal feature are assigned according to the environmental complexity; the weights of each modal feature include: infrared feature weight, polarization feature weight, and visible light feature weight; The temperature difference features, material contour features, and color texture features are fused and reconstructed using the modal feature weights to obtain the fused image data.

3. The intelligent monitoring method for safe production in hazardous work spaces according to claim 2, characterized in that, The calculation of environmental complexity based on the environmental parameter data, and the allocation of weights for each modal feature according to the environmental complexity, include: The dust concentration, light intensity, and smoke concentration were normalized to obtain normalized dust concentration, normalized light intensity, and normalized smoke concentration, respectively. The environmental complexity is obtained by multiplying the normalized dust concentration, normalized light intensity, and normalized smoke concentration by their respective preset weighting coefficients and then summing the results.

4. The intelligent monitoring method for safe production in hazardous work spaces according to claim 2, characterized in that, The calculation of environmental complexity based on the environmental parameter data, and the allocation of weights for each modal feature according to the environmental complexity, include: The infrared feature weights and polarization feature weights are determined based on preset base values ​​and the environmental complexity. The weights of visible light features are determined based on environmental complexity.

5. The intelligent monitoring method for safe production in hazardous work spaces according to claim 2, characterized in that, The method of fusing and reconstructing the temperature difference features, material contour features, and color texture features using the modal feature weights to obtain the fused image data includes: The temperature difference features, material contour features, and color texture features are multiplied by the corresponding modal feature weights, and then the product results are superimposed to obtain the pixel values ​​of each pixel in the fused image data, thus forming complete fused image data.

6. The intelligent monitoring method for safe production in hazardous work spaces according to claim 1, characterized in that, The target recognition processing based on the fused image data to obtain personnel behavior feature data and equipment operating status data includes: The environmental complexity calculated based on the environmental parameter data generates the recognition confidence threshold and the anchor frame size of the target recognition model; The target recognition model is adjusted based on the anchor frame size and the confidence threshold to generate a first recognition model; The first target recognition model is used to identify the fused image data to obtain the personnel behavior feature data and equipment operation status data.

7. The intelligent monitoring method for safe production in hazardous work spaces according to claim 1, characterized in that, The risk assessment calculation, which combines the personnel behavior data, equipment operating status data, and environmental parameter data, yields a comprehensive risk value including: Construct a set of personnel behavioral characteristics, which includes four parameters: personnel protective equipment wearing status, personnel entry status into risk areas, normalized result of personnel stay time, and personnel operation standardization. An environmental parameter set is constructed, which includes four parameters: normalized gas concentration, normalized temperature and humidity deviation, normalized dust concentration, and normalized light intensity. Assign corresponding weight coefficients to each parameter in the set of personnel behavioral characteristics and each parameter in the set of environmental parameters, and calculate the comprehensive risk value based on the configured weight coefficients, the parameters in the set of personnel behavioral characteristics, and the environmental parameters.

8. The intelligent monitoring method for safe production in hazardous work spaces according to claim 7, characterized in that, Assign corresponding weight coefficients to each parameter in the personnel behavior characteristic set and each parameter in the environmental parameter set. Calculate the comprehensive risk value based on the configured weight coefficients, the parameters in the personnel behavior characteristic set, and the environmental parameters, including: The parameters in the set of personnel behavior characteristics are converted into inverse values. The converted inverse values ​​of personnel behavior characteristics are multiplied by the corresponding weight coefficients to obtain the risk contribution value of each behavior. The summation is then used to obtain the total behavior risk value. Multiply each parameter in the set of environmental parameters by its corresponding weight coefficient to obtain the contribution value of each environmental risk, and sum them to obtain the total environmental risk value; The total behavioral risk value is added to the total environmental risk value to obtain the comprehensive risk value.

9. The intelligent monitoring method for safe production in hazardous work spaces according to claim 1, characterized in that, The step of triggering the corresponding early warning response operation based on the comprehensive risk value to complete the intelligent monitoring of safe production in hazardous work spaces includes: If the comprehensive risk value is greater than or equal to the first preset risk threshold, a level one warning is triggered, and on-site voice risk prompts, personnel location push notifications, and emergency department linkage operations are simultaneously initiated. If the comprehensive risk value is greater than or equal to the second preset risk threshold and less than the first preset risk threshold, a level two warning is triggered, and a warning message is pushed to the terminal device of the person in charge at the site, triggering a verification instruction. If the overall risk value is less than the second preset risk threshold, maintain normal monitoring status and transmit all data within this monitoring cycle to the storage module.

10. An intelligent monitoring system for safe production in hazardous work spaces, characterized in that, The system includes: Multimodal data acquisition module: used to acquire multimodal image data and environmental parameter data of hazardous work spaces. The multimodal image data includes infrared thermal imaging images, polarized light images and visible light images. The environmental parameter data includes dust concentration, light intensity, smoke concentration, gas concentration, temperature and humidity parameters. Image fusion processing module: used to fuse the multimodal image data to obtain fused image data; The target recognition module is used to perform target recognition processing based on the fused image data to obtain personnel behavior feature data and equipment operating status data; The risk assessment module is used to combine the personnel behavior characteristic data, equipment operating status data, and environmental parameter data to perform risk assessment calculations and obtain a comprehensive risk value. The early warning response module is used to trigger corresponding early warning response operations based on the comprehensive risk value, thereby completing intelligent monitoring of safe production in hazardous work spaces.

Images