An operation monitoring system and method based on an intelligent safety helmet

Abstract

The present invention provides an operation monitoring system and method based on an intelligent safety helmet, which relates to the technical field of operation monitoring. The method includes: obtaining the configuration information of the intelligent safety helmet and the operator; collecting the operation site information for AI analysis, and obtaining the second score of the sub-region of the operation site according to the AI analysis result; uploading the operation site information and the AI analysis result to the edge processing device according to the first score, the second score and the data score of the sub-region; performing a first processing on the operation site information through the edge processing device to obtain a first processing result; performing processing through the cloud platform to obtain a second processing result; giving an early warning or an alarm according to the AI analysis result, the first processing result and / or the second processing result; through the application of multi-dimensional data fusion and analysis, sub-region scoring and comprehensive evaluation, intelligent early warning and alarm system and machine learning model, the comprehensive, real-time and intelligent monitoring of the operation site and the operator is realized.

Description

Technical Field

[0001] This application relates to the technical field of operation monitoring, and particularly to an operation monitoring system and method based on an intelligent safety helmet. Background Art

[0002] With the continuous expansion of the scale and increasing complexity of modern engineering construction, the demand for engineering safety has also been rising. Traditional safety management methods mainly rely on manual supervision, which has many limitations, such as being difficult to achieve comprehensive and real-time monitoring and prone to missing potential safety hazards. Therefore, a more efficient and intelligent safety management system is needed to address these challenges.

[0003] In recent years, the Internet of Things (IoT) and artificial intelligence technologies have developed rapidly, making intelligent engineering safety management possible. The IoT technology enables various devices and sensors to be interconnected, realizing real-time data collection and transmission; while artificial intelligence technology can deeply analyze and process these data to discover potential safety risks. The combination of these technologies provides a solid technical foundation for the research and development of an intelligent safety helmet monitoring system.

[0004] As a safety protection device integrating a variety of advanced technologies, intelligent safety helmets have gradually attracted attention in recent years. By embedding hardware devices such as sensors and wireless communication modules in the safety helmet, it realizes real-time monitoring of the physiological indicators, environmental factors, and working status of the wearer.

[0005] However, there are still some problems with the current intelligent safety helmets and operation monitoring systems on the market: Existing intelligent safety helmets may only have single or a few data collection capabilities, such as positioning or environmental monitoring, lacking comprehensive on-site video monitoring and health monitoring functions. Although some systems can collect a certain amount of data, they lack effective AI analysis means to process these data in real time and cannot detect potential safety hazards in a timely manner. The existing systems have a relatively simple warning mechanism, often only triggering alarms based on thresholds without fully considering the complexity of the operation site and the comprehensive evaluation of various factors. There is a lack of a clear data processing architecture. From edge devices to the cloud, there is no clear plan on how to classify and process different types of data. Therefore, in order to improve the safety management level of the operation site, it is necessary to develop an intelligent safety helmet operation monitoring method that can comprehensively collect information on the operation site, use AI technology for efficient analysis, and provide intelligent decision support based on a multi-level data processing architecture. Summary of the Invention

[0006] The purpose of this application is to provide an integrated and intelligent operation monitoring method. Through multi-dimensional data fusion and analysis, sub-region scoring and comprehensive evaluation, intelligent early warning and alarm systems, and the application of machine learning models, comprehensive, real-time, and intelligent monitoring of the operation site and operators is achieved.

[0007] The purpose of this application is achieved by the following technical solutions:

[0008] This application provides an operation monitoring method based on an intelligent safety helmet. The method includes:

[0009] S1. Obtain the configuration information of the intelligent safety helmet and the operator;

[0010] S2. Divide the operation site into multiple sub-regions. Collect operation site information through the intelligent safety helmet. The operation site data includes images, videos, and on-site data. Perform AI analysis on the on-site images and videos through the GPU unit to obtain the AI analysis results;

[0011] S3. Obtain the second score of the sub-region of the operation site according to the AI analysis results; Upload the operation site information and the AI analysis results to the edge processing device according to the first score, second score, and data score of the sub-region;

[0012] S4. Perform the first processing on the operation site information through the edge processing device to obtain the first processing result;

[0013] S5. Upload the first processing result to the cloud platform and perform processing through the cloud platform to obtain the second processing result; The second processing result includes: the first danger probability of the sub-region, the second danger probability of the personnel, and the comprehensive danger probability;

[0014] S6. Perform early warning or alarm according to the AI analysis results, the first processing result, and / or the second processing result.

[0015] Preferably, the S1 includes:

[0016] Determine the basic information of the intelligent safety helmet. The basic information includes the serial number or unique identification number;

[0017] Configure the intelligent safety helmet according to the operation requirements, including setting communication parameters, safety areas, and SOS calling functions;

[0018] Obtain the information of the operator; The information includes identity information, position information, and basic health information;

[0019] Associate and match the operator information with the intelligent safety helmet; Obtain the permission for the operator to enter the corresponding operation area according to the matching result.

[0020] Preferably, the intelligent safety helmet includes a positioning unit, a video acquisition unit, a GPU unit, an environment acquisition unit, and a health acquisition unit, and the S2 includes:

[0021] Obtain the positioning data of the operator through the positioning unit;

[0022] The video acquisition unit is connected to the GPU unit, and the GPU unit performs AI analysis on the on-site images and videos collected by the video acquisition unit to obtain AI analysis results;

[0023] Obtain the environmental data of the operation site through the environment acquisition unit, and the environmental data includes temperature, humidity, and harmful gas concentration;

[0024] Obtain the health data of the operator through the health acquisition unit, and the health data includes body temperature, heart rate, and blood pressure.

[0025] Preferably, the S3 includes:

[0026] Score each sub-region according to the complexity and danger of the operation content to obtain the first score;

[0027] Obtain the second score of the operation site of each sub-region according to the AI analysis result;

[0028] The second score is:

[0029] ;

[0030] where Pj2 is the current second score of the jth sub-region; is the risk coefficient score obtained by the ith intelligent safety helmet in the jth sub-region according to the current AI analysis result;

[0031] Obtain the data comprehensive score according to the first score, the second score, and the data score;

[0032] where the comprehensive score is:

[0033] ;

[0034] ;

[0035] $Pzkv$ is the comprehensive score of the $v$-th group of data of the $k$-th safety helmet; $Pskv$ is the score of the $v$-th group of data of the $k$-th safety helmet; $Pj1$ is the first score of the $k$-th safety helmet in the $j$-th sub-region where it is currently located; $Pj2$ is the second score of the $k$-th safety helmet in the $j$-th sub-region where it is currently located; $Pskv$, $Pj1$, and $Pj2$ are all greater than 0; $x1$ and $x2$ are both the first weight coefficients; $Sskv$ is the real-time score of the $v$-th group of data of the $k$-th safety helmet; $Bskv$ is the score of the change in the $v$-th group of data of the $k$-th safety helmet compared with the previous sampling data.

[0036] According to the comprehensive data score, determine the transmission order of the data collected by the intelligent safety helmet; the higher the score, the higher the priority for transmission.

[0037] According to the transmission order, upload the data collected by the intelligent safety helmet to the edge processing device.

[0038] Preferably, the step $S4$ includes:

[0039] Correlate and correspond the environmental data, the health data of the operating personnel, and the positioning data.

[0040] Integrate the video capture results and AI analysis results of different safety helmets in the same sub-region and the same sampling time interval; the integration includes extracting the key frames of each safety helmet video stream, and performing clustering analysis on each key frame through a clustering algorithm, and removing duplicate frames according to the analysis results.

[0041] Perform statistical analysis on the environmental sampling data of the same type of intelligent safety helmets in the same sub-region within the same sampling time to obtain the mean, minimum value, and maximum value of each type of data.

[0042] Perform statistical analysis on the health sampling data of each safety helmet.

[0043] Preferably, the step $S5$ includes:

[0044] Based on the real-time video data and environmental data, and using the first machine learning model, predict the first risk probability of the sub-region.

[0045] Based on the real-time health data, and using the second machine learning model, predict the second risk probability.

[0046] Obtain the comprehensive risk probability through the first risk probability and the second risk probability.

[0047] Based on the first risk probability, the second risk probability, and / or the comprehensive risk probability, give an early warning of personnel safety.

[0048] Preferably, the input of the first machine learning model includes: the features of real-time video data, the mean of each environmental data sampled by multiple safety helmets in the sub-region currently, the maximum value of each environmental data sampled by multiple safety helmets in the sub-region currently, the minimum value of each environmental data sampled by multiple safety helmets in the sub-region currently, and the mean change of each environmental data sampled by multiple safety helmets in the sub-region in the most recent multiple samplings;

[0049] The input of the second machine learning model includes: each parameter sampled for the current health of each safety helmet wearer and the change value of each parameter in multiple samplings.

[0050] Preferably, obtaining the comprehensive risk probability through the first risk probability and the second risk probability includes:

[0051] The comprehensive risk probability is obtained by the following formula:

[0052] ;

[0053] where Fau is the comprehensive risk probability of the u-th person; f1 is the first risk probability of the sub-region where the u-th person is currently located; and fu2 is the second risk probability of the u-th person.

[0054] Preferably, S6 includes:

[0055] Obtaining a first anomaly according to the AI analysis result and giving a first alarm according to the first anomaly;

[0056] If the first processing result indicates an anomaly in the health of the person, then give a second alarm;

[0057] If any one of the first risk probability, the second risk probability, or the comprehensive risk probability exceeds the corresponding probability threshold, then give a warning about the safety of the person.

[0058] This application provides a system for implementing the operation monitoring method based on intelligent safety helmets described in this application. The system includes:

[0059] A configuration module, used to obtain the configuration information of intelligent safety helmets and operating personnel;

[0060] An acquisition module, used to divide the operation site into multiple sub-regions, collect operation site information through intelligent safety helmets. The operation site data includes images, videos, and on-site data, and perform AI analysis on the on-site images and videos through a GPU unit to obtain an AI analysis result;

[0061] A first transmission module, used to obtain a second score of the operation site sub-region according to the AI analysis result; and upload the operation site information and the AI analysis result to the edge processing device according to the first score, the second score, and the data score of the sub-region;

[0062] The first processing module is configured to perform first processing on the job site information through an edge processing device to obtain a first processing result;

[0063] The second processing module is configured to upload the first processing result to a cloud platform for processing to obtain a second processing result; the second processing result includes: the first danger probability of a sub-region, the second danger probability of a person, and the comprehensive danger probability;

[0064] The warning module is configured to issue a warning or an alarm according to the AI analysis result, the first processing result, and / or the second processing result.

[0065] The beneficial effects of the present invention include: The present application proposes a job monitoring method based on an intelligent safety helmet. The intelligent safety helmet integrates multiple sensors (such as a positioning unit, a video acquisition unit, an environment acquisition unit, and a health acquisition unit) and a GPU unit, and can collect multi-dimensional data (images, videos, environment data, health data, etc.) of the job site in real time, and achieve comprehensive monitoring of the job site through AI analysis; through the GPU unit, AI analysis is performed on the on-site images and videos, combined with the environment data and health data, to achieve in-depth insight into the state of the job site and the job personnel. At the same time, through the collaborative work of the edge processing device and the cloud platform, the efficiency and accuracy of data processing are further improved; a sub-region scoring mechanism is introduced, each sub-region is scored according to the complexity and danger of the job content, and a second score is obtained in combination with the AI analysis result. Finally, the priority of data transmission is determined through data comprehensive scoring. This method can ensure the priority transmission of important data and improve the response speed of the monitoring system; based on the AI analysis result, the first processing result, and the second processing result, the present application constructs a set of intelligent warning and alarm systems. The system can monitor the state of the job site and the job personnel in real time. Once an abnormal situation is found, a warning or an alarm is immediately issued, effectively reducing the probability of accidents; the present application uses a machine learning model to perform predictive analysis on real-time video data, environment data, and health data to obtain the first danger probability of a sub-region and the second danger probability of a person, and then calculates the comprehensive danger probability; improving the accuracy and timeliness of warning and alarm. The present application also involves the acquisition of configuration information and permission management of intelligent safety helmets and job personnel, ensuring that only job personnel with corresponding permissions can enter the corresponding job area, enhancing the safety and management efficiency of the job site. Description of the Drawings

[0066] Figure 1 It is a schematic diagram of a job monitoring method based on an intelligent safety helmet provided by an embodiment of the present application. Detailed Embodiments

[0067] Next, in combination with the accompanying drawings and specific embodiments, the present application will be further described. It should be noted that, on the premise of no conflict, the following-described embodiments or technical features can be arbitrarily combined to form new embodiments.

[0068] Referring to Figure 1 , an embodiment of the present application provides an operation monitoring method based on an intelligent safety helmet. The method includes:

[0069] S1. Obtain the configuration information of the intelligent safety helmet and the operator;

[0070] S2. Divide the operation site into multiple sub-regions, collect operation site information through the intelligent safety helmet. The operation site data includes images, videos, and on-site data. Perform AI analysis on the on-site images and videos through the GPU unit to obtain the AI analysis result; the on-site data includes operator positioning data, environmental data, and operator health data;

[0071] S3. Obtain the second score of the sub-region of the operation site according to the AI analysis result; upload the operation site information and the AI analysis result to the edge processing device according to the first score, the second score, and the data score of the sub-region;

[0072] S4. Perform the first processing on the operation site information through the edge processing device to obtain the first processing result;

[0073] S5. Upload the first processing result to the cloud platform, and perform processing through the cloud platform to obtain the second processing result; the second processing result includes: the first danger probability of the sub-region, the second danger probability of the personnel, and the comprehensive danger probability;

[0074] S6. Give an early warning or alarm according to the AI analysis result, the first processing result, and / or the second processing result.

[0075] The working principle and effect of the above technical solution are as follows:

[0076] First of all, the system needs to obtain the configuration information of the intelligent safety helmet and the operator. This information may include the model of the intelligent safety helmet, functional parameters, the identity information of the operator, skill level, and special safety requirements, etc. These configuration information provide a basis for subsequent data processing and risk assessment.

[0077] The operation site is divided into multiple sub-regions for more precise monitoring and management. Through the sensors and cameras built in the intelligent safety helmet, the images, videos, and various on-site data of the operation site are collected in real time (such as operator positioning data, environmental data such as temperature, humidity, gas concentration, and operator health data such as heart rate, body temperature, etc.).

[0078] Utilize the powerful computing power of GPU units to perform AI analysis on the collected images and videos, identify potential safety hazards or abnormal situations, such as not wearing safety equipment, illegal operations, dangerous environments, etc., and generate AI analysis results.

[0079] Based on the AI analysis results, conduct a second scoring for sub-regions of the operation site; at the same time, combine the existing first scoring and data scoring (direct assessment of on-site data) to comprehensively evaluate the safety status of the operation site.

[0080] Upload the operation site information, AI analysis results, and scoring data to the edge processing device for further processing and analysis.

[0081] In the edge processing device, perform preliminary processing and analysis on the uploaded data, such as data cleaning, feature extraction, etc., to reduce the processing pressure on the cloud platform and speed up the response speed. The processing results (the first processing results) are then uploaded to the cloud platform for more advanced analysis.

[0082] In the cloud platform, conduct in-depth analysis and modeling on the received data, and calculate key indicators such as the first danger probability of the sub-region, the second danger probability of personnel, and the comprehensive danger probability. These indicators comprehensively reflect the safety risk level of the operation site.

[0083] Based on the AI analysis results, the first processing results, and / or the second processing results, the system can automatically identify potential safety risks and trigger a warning or alarm mechanism according to preset thresholds or rules. Once the warning or alarm is triggered, the system can send safety tips or emergency evacuation instructions to the operation personnel through intelligent safety helmets or other communication devices, and at the same time notify the safety management personnel for timely handling.

[0084] In summary, this method uses intelligent safety helmets to collect on-site operation data in real time, combines AI analysis and multi-level processing mechanisms to achieve real-time monitoring and early warning of safety risks at the operation site, and effectively improves the efficiency and accuracy of operation safety management.

[0085] In some embodiments, the S1 includes:

[0086] Determine the basic information of the intelligent safety helmet, and the basic information includes the serial number or unique identification number;

[0087] Configure the intelligent safety helmet according to the operation requirements, including setting communication parameters, safety areas, and SOS calling functions;

[0088] Obtain the information of the operation personnel; the information includes identity information, position information, and basic health information;

[0089] Associate the information of the operator with the intelligent safety helmet; obtain the permission for the operator to enter the corresponding working area according to the matching result.

[0090] The working principle and effect of the above technical solution are as follows:

[0091] Determine the basic information of the intelligent safety helmet to ensure that each intelligent safety helmet has a unique identifier, such as a serial number or a unique identification number; this is crucial for subsequent tracking, management, and data analysis.

[0092] Configure the intelligent safety helmet according to the operation requirements: In this step, the intelligent safety helmet is personalized configured according to specific working environments and requirements. This includes setting communication parameters (such as wireless communication frequency bands, data transmission rates, etc.), defining safety areas (limiting the activity range of operators), and enabling the SOS call function (to quickly send out distress signals in case of emergencies). These configurations ensure that the intelligent safety helmet can meet the needs of different working scenarios.

[0093] Obtain the information of the operator: This step collects the key information of the operator, such as identity information, position information, and basic health information. These information are crucial for verifying the identity of the operator, clarifying their responsibilities, and paying attention to their health status.

[0094] Associate and match the operator information with the intelligent safety helmet: This step is the key to connecting the operator and the intelligent safety helmet. By some means (such as scanning a QR code, RFID identification, etc.), the personal information of the operator is associated with the unique identification number of the intelligent safety helmet, and the system can know which safety helmet is worn by which operator, so as to track and record the activity data of the operator.

[0095] After completing the association and matching, the system will automatically assign the permission for the operator to enter the corresponding working area according to the operator's position information and safety regulations. This permission management mechanism helps to ensure that only operators with the corresponding qualifications and permissions can enter specific working areas, thus improving the safety and management efficiency of the working site.

[0096] In some embodiments, the intelligent safety helmet includes a positioning unit, a video acquisition unit, a GPU unit, an environment acquisition unit, and a health acquisition unit, and the S2 includes:

[0097] Divide the working site into multiple sub-areas;

[0098] Obtain the positioning data of the operator through the positioning unit; the positioning unit includes but is not limited to GPS and Beidou satellites;

[0099] The video acquisition unit is connected to the GPU unit, and the video acquisition unit is used to acquire on-site images and videos. The GPU unit performs AI analysis on the on-site images and videos through an AI model to obtain AI analysis results. The AI ​​analysis results include:

[0100] Identify whether the operators are wearing necessary protective equipment such as safety helmets and reflective clothing, and whether they comply with safety operating procedures, such as whether they are operating in violation of regulations in high-risk areas;

[0101] Evaluate the quality of operators' work, such as whether the operation is standardized and whether it meets the established operating standards;

[0102] Detect whether there are abnormal behaviors at the work site, such as gathering of people, entering restricted areas, etc.;

[0103] Acquire environmental data of the work site through an environmental collection unit, wherein the environmental data includes temperature, humidity, and concentration of harmful gases;

[0104] The health data of the operator is obtained through the health collection unit, and the health data includes body temperature, heart rate and blood pressure.

[0105] The principle and effect of the above technical solution are:

[0106] Divide the work site into multiple sub-areas; ensure that each sub-area can be accurately identified and defined; the specific division method is not specified here, for example:

[0107] By function:

[0108] According to the different functional requirements of the work site, the site is divided into sub-areas such as construction work area, auxiliary work area, material storage area, office area, and living area.

[0109] According to the construction content:

[0110] The site is divided into different construction sections or construction layers according to the different construction contents. For example, in the construction of multi-storey buildings, each floor or several floors can be divided into a construction section; in large infrastructure projects, different construction areas can be divided according to different construction tasks (such as foundation treatment, main structure construction, decoration and renovation, etc.).

[0111] It can also be based on geographic information system (GIS) data or other spatial data to ensure that each sub-area can be accurately identified and defined. For example, in urban construction site monitoring, the entire site can be divided into sub-areas such as different floors and different building units.

[0112] By receiving satellite signals or wireless network signals, the positioning unit calculates the operator's latitude, longitude, altitude and other location information, and updates these data in real time. These data are crucial for monitoring the operator's movement trajectory and determining whether they have entered a dangerous area.

[0113] The video acquisition unit captures the real-time images of the scene through the camera and transmits the data to the GPU unit;

[0114] AI analysis: GPU units have built-in or external AI models to perform in-depth analysis of received images and videos. The analysis content includes but is not limited to:

[0115] Identify whether workers are wearing necessary safety equipment (such as helmets and reflective clothing).

[0116] Monitor whether operators comply with safety operating procedures, such as whether they operate in violation of regulations in high-risk areas.

[0117] Evaluate the work quality of operators to determine whether their operations are standardized and meet established standards.

[0118] Detect whether there are abnormal behaviors at the work site, such as gathering of people, entering restricted areas, etc.

[0119] Result output: After the AI ​​analysis is completed, the results are output in a structured form for subsequent processing.

[0120] Through built-in sensors (such as temperature and humidity sensors, gas concentration sensors), the environmental acquisition unit collects environmental data in real time and transmits this data to the processing system. This data is of great significance for assessing the safety of the working environment and predicting potential dangers.

[0121] The health collection unit pays attention to the physical health of the workers and monitors their key indicators such as body temperature, heart rate, and blood pressure. Through contact or non-contact sensors (such as infrared body temperature sensors, heart rate belts, blood pressure monitors, etc.), the health collection unit collects the health data of workers in real time. These data help to detect the physical discomfort or abnormal conditions of the workers in a timely manner, so as to take corresponding preventive measures. The positioning data, video analysis results, environmental data, and health data are integrated for comprehensive analysis and processing.

[0122] Aggregate the data collected by each unit to the processing system (such as edge processing devices or cloud platforms).

[0123] The processing system uses preset algorithms and models to comprehensively analyze this data, evaluate the safety status of the operation site, the operation quality of the operators, and their health status. According to the analysis results, the system can trigger an early warning or alarm mechanism to send safety tips or emergency evacuation instructions to the operators and management personnel; at the same time, it can also provide data support for subsequent safety management decisions.

[0124] In summary, the intelligent safety helmets in these embodiments achieve comprehensive monitoring and analysis of the operation site and operators by integrating multiple sensors and units, providing strong support for improving the operation safety management level.

[0125] In some embodiments, S3 includes:

[0126] Score each sub-region according to the complexity and danger of the operation content to obtain the first score;

[0127] Obtain the second score of the operation site in each sub-region according to the AI analysis results;

[0128] The second score is:

[0129] ;

[0130] where Pj2 is the current second score of the j-th sub-region; is the risk coefficient score obtained by the i-th intelligent safety helmet in the j-th sub-region according to the current AI analysis results;

[0131] Obtain the data comprehensive score according to the first score, the second score, and the data score;

[0132] where the comprehensive score is:

[0133] ;

[0134] ;

[0135] Pzkv is the comprehensive score of the k-th safety helmet for the v-th group of data; Pskv is the score of the k-th safety helmet for the v-th group of data; Pj1 is the first score of the j-th sub-region where the k-th safety helmet is currently located; Pj2 is the second score of the j-th sub-region where the k-th safety helmet is currently located; Pskv, Pj1, and Pj2 are all greater than 0; x1 and x2 are both the first weight coefficients; Sskv is the real-time score of the k-th safety helmet for the v-th group of data; Bskv is the score of the change in the k-th safety helmet for the v-th group of data compared with the previous sampling data;

[0136] Determine the transmission order of the data collected by the intelligent safety helmet according to the data comprehensive score; the higher the score, the higher the priority for transmission;

[0137] Upload the data collected by the intelligent safety helmet to the edge processing device according to the transmission order.

[0138] The working principle and effect of the above technical solution are as follows: perform a first score on each sub-region according to the complexity and danger of the operation content; this score can be obtained based on prior knowledge or expert evaluation.

[0139] Through the AI analysis results, calculate the average of the risk factor scores provided by all intelligent safety helmets in each sub-region to obtain the real-time safety score of the sub-region; this score dynamically reflects the current safety status of the sub-region.

[0140] Evaluate the real-time nature of the v-th group of data of the k-th intelligent safety helmet, that is, the freshness and timeliness of the data.

[0141] Evaluate the degree of change between the v-th group of data of the k-th intelligent safety helmet and the previous sampled data, which reflects the dynamics and importance of the data.

[0142] Combine the real-time score, change score, and the first and second scores of the sub-region (adjusted by weight coefficients x1 and x2) to calculate the comprehensive score of each group of data of each intelligent safety helmet. This score comprehensively considers multiple dimensions of the data, including timeliness, variability, and the safety status of the sub-region.

[0143] Sort the data collected by all intelligent safety helmets according to their comprehensive scores, and the data with higher scores is transmitted first. Upload the data collected by the intelligent safety helmet to the edge processing device according to the determined transmission order. This mechanism ensures that key, urgent, or important data can be processed first, improving the response speed and efficiency of the entire system.

[0144] Step S3 realizes a comprehensive evaluation of the data collected by the intelligent safety helmet by constructing a set of comprehensive scoring mechanisms, taking into account multiple factors such as the inherent risks of the sub-region, real-time safety status, and the real-time nature and variability of the data. By preferentially transmitting data with higher scores, that is, data that is more likely to contain important or urgent information, the overall efficiency of data transmission can be significantly improved. This helps to ensure that key information can be processed and analyzed faster, thus responding to potential safety risks in a timely manner; improving the safety monitoring efficiency and response speed at the operation site not only helps to detect and respond to potential safety risks in a timely manner, but also provides strong data support for subsequent decision-making analysis and safety management; by introducing real-time scores and data change scores, the system can evaluate the timeliness and importance of the data. This helps to screen out more valuable data for transmission and processing, thereby improving the overall quality and reliability of the data.

[0145] In some embodiments, the S4 includes:

[0146] Associate and correspond environmental data, workers' health data, and positioning data;

[0147] Integrate the video capture results and AI analysis results of different safety helmets in the same sub-region and the same sampling time interval; The integration includes extracting the key frames of each safety helmet video stream, and performing clustering analysis on each key frame through a clustering algorithm, and removing duplicate frames according to the analysis results;

[0148] Statistically analyze the same type of environmental sampling data of intelligent safety helmets in the same sub-region within the same sampling time, and obtain the mean, minimum, and maximum values of each type of data;

[0149] Statistically analyze the health sampling numbers of each safety helmet.

[0150] The working principle and effect of the above technical solution are as follows: First, perform timestamp matching and association on the collected environmental data (such as temperature, humidity, harmful gas concentration, etc.), workers' health data (such as body temperature, heart rate, blood pressure, etc.), and positioning data (such as GPS coordinates); Each group of data can be corresponding to a specific time, location, and person, providing a basis for subsequent analysis.

[0151] For the video capture results of different safety helmets in the same sub-region and the same sampling time interval, first extract the key frames in their respective video streams; Key frames are the frames in the video that can represent its main content, usually containing more information and less redundancy.

[0152] Then, perform clustering analysis on all the extracted key frames using a clustering algorithm; The clustering algorithm will divide them into different groups according to the similarity between the key frames (such as features like color, texture, shape, etc.). By comparing the key frames within each group, duplicate or highly similar frames can be identified and removed to reduce the burden of data storage and processing.

[0153] At the same time, integrate the AI analysis results of each safety helmet (such as whether the safety helmet is worn, whether safety regulations are complied with, job quality assessment, etc.). These results will be associated with the video key frames to provide more comprehensive information in subsequent analysis and display.

[0154] For the same type of environmental data (such as temperature, humidity, etc.) collected by intelligent safety helmets in the same sub-region within the same sampling time interval, perform statistical analysis; Calculate the mean, minimum, and maximum values of each type of data to understand the environmental conditions and their fluctuation ranges in this sub-region during this time period.

[0155] Statistically analyze the health data collected from each safety helmet. This includes calculating statistics such as the mean, standard deviation, etc. of the health indicators (such as body temperature, heart rate, blood pressure) of each worker to evaluate their overall health status and change trends. At the same time, outlier detection can also be carried out to timely detect possible health problems or abnormal conditions.

[0156] In step S4, through means such as data association, video integration, environmental data statistical analysis, and health data statistical analysis, the comprehensive processing and analysis of multi-source data at the work site are realized. By integrating the video capture results of different safety helmets within the same sub-region and the same sampling time interval, and using the clustering algorithm to remove duplicate frames, the redundant information in the data is effectively reduced. This not only saves storage space but also improves the efficiency of data processing and transmission. By integrating data from different sources, a more comprehensive and in-depth understanding of the work site can be formed, which helps to timely discover and solve potential safety problems. Statistically analyze the environmental sampling data of the same type of intelligent safety helmets within the same sub-region at the same sampling time to obtain statistical indicators such as the mean, minimum value, and maximum value, which helps to more accurately understand the change trend and distribution of the environmental conditions; directly transmitting the original environmental sampling data to the cloud platform will consume a large amount of bandwidth and storage resources. By first performing statistical analysis on local or edge devices and only transmitting the statistical results (such as the mean, minimum value, maximum value) instead of all the original data, the data transmission volume can be significantly reduced, the data transmission efficiency can be improved, and the processing burden and operating costs of the cloud platform can be reduced. Performing statistical analysis on edge devices or locally can obtain an overview of the environmental conditions faster, which helps the system to respond to environmental changes in a timely manner and take necessary adjustment measures. For example, when it is detected that the temperature or harmful gas concentration exceeds the safety range, the system can immediately trigger an alarm and notify relevant personnel to take corresponding measures. Statistical indicators, as a highly generalized description of the environmental conditions, provide strong support for subsequent data analysis and decision-making. Based on these statistical indicators, the system can further perform advanced analysis such as trend prediction and outlier detection to provide a more scientific and comprehensive decision-making basis for management.

[0157] In some embodiments, the S5 includes:

[0158] Predict the first risk probability of the sub-region based on the real-time video data and environmental data using the first machine learning model;

[0159] Predict the second risk probability based on the real-time health data using the second machine learning model;

[0160] Obtain the comprehensive risk probability based on the first risk probability and the second risk probability;

[0161] Warn of personnel safety according to the first risk probability, the second risk probability, and / or the comprehensive risk probability.

[0162] In some embodiments, the inputs of the first machine learning model include: features of real-time video data, the mean values of various environmental data sampled currently by multiple safety helmets in a sub-region, the maximum values of various environmental data sampled currently by multiple safety helmets in a sub-region, the minimum values of various environmental data sampled currently by multiple safety helmets in a sub-region, and the mean change amount of the most recent multiple samplings of various environmental data sampled currently by multiple safety helmets in a sub-region;

[0163] The inputs of the second machine learning model include: various parameters sampled currently for the health of each safety helmet wearer and the change values of multiple samplings of each parameter.

[0164] The working principle of the above technical solution is as follows: The system first collects features of real-time video data (such as motion patterns in images, object recognition results, etc.), and environmental data sampled currently by multiple intelligent safety helmets in a sub-region (such as temperature, humidity, concentration of harmful gases, etc.). For the environmental data, not only the sampled values (mean, maximum, minimum) of multiple safety helmets currently are considered, but also the mean change amount of the most recent multiple samplings of these data is considered to capture the dynamic changes of environmental conditions. The above data is input into the first machine learning model. This model has been trained to be able to identify abnormal patterns in environmental data and unsafe behaviors or states in video data, so as to predict the first risk probability of the sub-region. This probability reflects the potential risk level evaluated only based on environmental and video data.

[0165] The system simultaneously collects the current health data of the operators collected by each intelligent safety helmet (such as body temperature, heart rate, blood pressure, etc.), and the change values of multiple samplings of these health parameters. These data reflect the real-time health status and change trends of the personnel.

[0166] The health data is input into the second machine learning model. This model has been trained to be able to identify abnormal values or abnormal change trends in health parameters, so as to predict the second risk probability. This probability reflects the potential risk level evaluated only based on health data, such as possible health problems or excessive fatigue of the personnel.

[0167] According to the first risk probability and the second risk probability, through a certain weight or algorithm (such as weighted average, maximum value selection, etc.), a comprehensive risk probability is calculated; this probability combines the risk factors in both the environmental and health aspects and provides a more comprehensive assessment for personnel safety.

[0168] The system sets different warning levels and thresholds according to the first risk probability, the second risk probability, and / or the comprehensive risk probability. When any one of the risk probabilities or the comprehensive risk probability exceeds the set probability threshold, the system triggers the corresponding warning mechanism.

[0169] Early warning measures may include sending alert messages to relevant personnel, automatically adjusting the working environment (such as turning on ventilation equipment, adjusting temperature, etc.), suggesting that personnel take breaks or adopt other safety measures. These measures are aimed at timely reminding personnel to pay attention to potential dangers and taking measures to reduce risks.

[0170] In step S5, by combining real-time video data, environmental data, and health data, two machine learning models are used to predict the danger probabilities of sub-regions and personnel respectively, and comprehensive evaluation and early warning are carried out based on these prediction results. Through multi-dimensional and real-time risk assessment and intelligent prediction, the system can detect potential safety hazards earlier and issue early warning signals, thereby enhancing the timeliness and accuracy of safety early warning.

[0171] In some embodiments, the comprehensive danger probability is obtained through the first danger probability and the second danger probability; including:

[0172] The comprehensive danger probability is obtained through the following formula:

[0173] ;

[0174] where Fau is the comprehensive danger probability of the u-th person; f1 is the first danger probability of the sub-region where the u-th person is currently located; fu2 is the second danger probability of the u-th person, where .

[0175] The working principle of the above technical solution is: The formula is actually a variant of weighted average, and it is not a simple weighted average in the traditional sense. In this formula, each danger probability (f1 and fu2) is squared and then added together, and then divided by the sum of these two probabilities.

[0176] If any one of the values of f1 or fu2 is relatively high, then its contribution to the comprehensive danger probability will be more significant. This means that if the danger probability of the sub-region or the individual is relatively high, then the comprehensive danger probability will also be correspondingly higher.

[0177] The denominator is f1 + fu2, which ensures that even if one of the values of f1 or fu2 is zero (i.e., there is no danger), the denominator will not be zero, thus avoiding the error of division by zero. At the same time, the denominator also plays a role in normalization, controlling the value range of the comprehensive danger probability Fau.

[0178] First of all, the system needs to obtain the first danger probability f1 of the sub-region where the u-th person is currently located and the second danger probability fu2 of this person. These probability values are usually calculated through machine learning models or other risk assessment methods.

[0179] Then, the system calculates the comprehensive risk probability Fau of the u-th person using the above formula. This calculation process takes into account the overall risk of the sub-region and the risk of the individual's health condition, and combines them in a certain way.

[0180] Generally speaking, this formula provides a method for calculating the comprehensive risk probability by combining the risk probabilities at the sub-region and individual levels. This method not only considers the safety of the overall environment but also pays attention to the individual's health condition, which helps to more comprehensively evaluate the safety risk.

[0181] The effect of the above technical solution is that due to the square operation amplifying the influence of larger values, this formula can more strongly reflect the high-risk factors in the sub-region or the individual. If any value in f1 or fu2 is relatively high, then its contribution to the comprehensive risk probability will be more significant and thus more likely to attract attention.

[0182] The formula takes into account both the first risk probability (f1) of the sub-region and the second risk probability (fu2) of the individual, and calculates the comprehensive risk probability by combining them, achieving a balanced consideration of the overall risk of the sub-region and the individual risk; it helps to more comprehensively evaluate the safety risk and avoid only focusing on a single level while ignoring other important factors.

[0183] In some embodiments, S6 includes:

[0184] Obtain the first anomaly according to the AI analysis result and issue the first alarm according to the first anomaly;

[0185] If the first processing result indicates an anomaly in the person's health, then issue the second alarm;

[0186] If any one of the first risk probability, the second risk probability, or the comprehensive risk probability exceeds the corresponding probability threshold, then give a warning about the person's safety.

[0187] The working principle of the above technical solution is:

[0188] First, the AI algorithm integrated with the GPU performs real-time analysis on the collected data;

[0189] The AI algorithm uses machine learning or deep learning techniques to identify abnormal situations that do not conform to the normal state, namely the "first anomaly", from the massive data through methods such as pattern recognition and anomaly detection.

[0190] AI may identify fatigue operation or improper operation as the first anomaly by analyzing the worker's motion pattern.

[0191] Once the AI analysis determines the existence of a first anomaly, the system immediately triggers the first alarm mechanism. The alarm form can be a sound alarm, a text message notification, an email reminder, or directly display a warning message on the user interface.

[0192] If the first processing result indicates a health anomaly of a person, the system will issue a second alarm. The second alarm is usually more urgent and specific than the first alarm because it is directly related to the life safety of the person. For example, if the second processing result determines that a certain patient has an abnormal heart rate, the second alarm will be immediately triggered, which may include directly calling the emergency team, activating emergency medical equipment, etc.

[0193] The system will also calculate the first risk probability (based on the first anomaly), the second risk probability (based on the confirmation of health anomaly), and the comprehensive risk probability (considering all relevant factors). If any of these probabilities exceeds a preset threshold, the system will issue a warning about the safety of the person. The warning may include more detailed risk tips, safety suggestions, or automatically taken safety measures.

[0194] For example, in a chemical plant, if the AI detects a toxic gas leak and calculates that the first risk probability is extremely high, the system will immediately issue a warning and may automatically activate the ventilation system, shut down relevant equipment, and notify all personnel to evacuate.

[0195] In summary, through AI analysis, multi-level alarms, and a warning mechanism based on risk probability, this workflow effectively improves the efficiency and accuracy of personnel health and safety monitoring.

[0196] An embodiment of the present application provides a system for the operation monitoring method based on an intelligent safety helmet described in the embodiments of the present application. The system includes:

[0197] A configuration module for obtaining the configuration information of the intelligent safety helmet and the operating personnel;

[0198] An acquisition module for dividing the operation site into multiple sub-regions, collecting operation site information through the intelligent safety helmet. The operation site data includes images, videos, and on-site data. Through the GPU unit, AI analysis is performed on the on-site images and videos to obtain the AI analysis result; the on-site data includes the positioning data of the operating personnel, environmental data, and the health data of the operating personnel;

[0199] A first transmission module for obtaining the second score of the sub-region of the operation site according to the AI analysis result; uploading the operation site information and the AI analysis result to the edge processing device according to the first score, the second score, and the data score of the sub-region;

[0200] A first processing module for performing a first processing on the operation site information through the edge processing device to obtain a first processing result;

[0201] A second processing module, configured to upload the first processing result to a cloud platform for processing to obtain a second processing result; the second processing result includes: a first risk probability of a sub-region, a second risk probability of a person, and a comprehensive risk probability;

[0202] An early warning module, configured to give an early warning or an alarm according to the AI analysis result, the first processing result, and / or the second processing result.

[0203] The working principle and effect of the above technical solution are the same as those in the method embodiment of the present application, and will not be elaborated here.

[0204] The present application is described from the perspectives of purpose of use, effectiveness, progress, and novelty, and has met the functional enhancement and use requirements emphasized by the patent law. The above description and the accompanying drawings of the present application are only preferred embodiments of the present application, and do not limit the present application thereto. Therefore, all those that are similar or identical to the structure, device, features, etc. of the present application, that is, all equivalent replacements or modifications made according to the scope of the patent application of the present application, shall fall within the scope of protection of the patent application of the present application.

Claims

1. A method for monitoring work based on a smart helmet, characterized in that: The method comprises: S1. Obtain the configuration information of the smart helmet and the operator; S2. Divide the work site into multiple sub-areas, collect work site information through the smart safety helmet, the work site information includes images, videos and field data, and perform AI analysis on the field images and videos through the GPU unit to obtain AI analysis results; S3. Obtain a second score of the sub-area of ​​the work site according to the AI ​​analysis result; upload the work site information and the AI ​​analysis result to the edge processing device according to the first score, the second score and the data score of the sub-area; S4. Performing a first process on the work site information by an edge processing device to obtain a first processing result; S5, uploading the first processing result to the cloud platform, processing through the cloud platform, and obtaining a second processing result; the second processing result includes: the first danger probability of the sub-area, the second danger probability of the personnel, and the comprehensive danger probability; S6. issuing an early warning or an alarm according to the AI ​​analysis result, the first processing result and / or the second processing result; The S3 includes: Each sub-area is scored based on the complexity and dangerousness of the work content, and a first score is obtained; Obtain a second score for each sub-area work site based on the AI ​​analysis results; The second rating is: ； Wherein, Pj2 is the current second score of the j-th sub-region; The risk factor score obtained for the i-th smart helmet in the j-th sub-area based on the current AI analysis results; According to the first score, the second score and the data score, a comprehensive data score is obtained; The comprehensive rating is: ； ； Pzkv is the comprehensive score of the k-th helmet's v-th group data; Pskv is the score of the k-th helmet's v-th group data; Pj1' is the first score of the k-th helmet's current j-th sub-region; Pj2' is the second score of the k-th helmet's current j-th sub-region; Pskv, Pj1', Pj2' are all greater than 0; x1 and x2 are both the first weight coefficients; Sskv is the real-time score of the k-th helmet's v-th group data; Bskv is the score of the change between the k-th helmet's v-th group data and the previous sampling data; According to the comprehensive data score, the transmission order of the data collected by the smart helmet is determined; the higher the score, the higher the priority of transmission; According to the transmission sequence, the data collected by the smart helmet is uploaded to the edge processing device.

2. The monitoring method according to claim 1, characterized in that: The S1 includes: Determine basic information of the smart helmet, the basic information including a serial number or a unique identification number; Configure the smart helmet according to the job requirements, including setting communication parameters, safety zones, and SOS call functions; Obtaining the information of the operator; the information includes identity information, position information and basic health information; The operator information is associated and matched with the smart safety helmet; the operator's permission to enter the corresponding work area is obtained based on the matching results.

3. The monitoring method according to claim 1, characterized in that: The smart helmet includes a positioning unit, a video acquisition unit, a GPU unit, an environment acquisition unit and a health acquisition unit, and S2 includes: Acquiring positioning data of the operator through the positioning unit; The video acquisition unit is connected to the GPU unit, and the video acquisition unit is used to acquire on-site images and videos. The GPU unit performs AI analysis on the on-site images and videos through an AI model to obtain AI analysis results. Acquire environmental data of the work site through an environmental collection unit, wherein the environmental data includes temperature, humidity, and concentration of harmful gases; The health data of the operator is obtained through the health collection unit, and the health data includes body temperature, heart rate and blood pressure.

4. The monitoring method according to claim 1, characterized in that: The S4 includes: Correlate environmental data, worker health data, and positioning data; Integrate the video acquisition results and AI analysis results of different helmets in the same sub-area and the same sampling time interval; the integration includes extracting key frames of each helmet video stream, performing cluster analysis on each key frame through a clustering algorithm, and removing duplicate frames according to the analysis results; Perform statistical analysis on the same type of environmental sampling data of smart helmets in the same sub-area during the same sampling time to obtain the mean, minimum and maximum values ​​of each type of data; The healthy sampling numbers of each helmet were statistically analyzed.

5. The monitoring method according to claim 1, characterized in that: The S5 includes: Predicting a first danger probability of a sub-area based on a first machine learning model using real-time video data and environmental data; Predicting a second risk probability based on a second machine learning model using real-time health data; Obtain a comprehensive risk probability through the first risk probability and the second risk probability; According to the first danger probability, the second danger probability and / or the comprehensive danger probability, an early warning is issued for personnel safety.

6. The monitoring method according to claim 5, characterized in that: The input of the first machine learning model includes: features of real-time video data, mean values ​​of environmental data currently sampled by multiple helmets in the sub-area, maximum values ​​of environmental data currently sampled by multiple helmets in the sub-area, minimum values ​​of environmental data currently sampled by multiple helmets in the sub-area, and mean changes of the most recent multiple samples of environmental data currently sampled by multiple helmets in the sub-area; The input of the second machine learning model includes: the current health sampling parameters of each helmet worker and the change values ​​of multiple samplings of each parameter.

7. The monitoring method according to claim 5, characterized in that: The method of obtaining a comprehensive risk probability through the first risk probability and the second risk probability comprises: The comprehensive hazard probability is obtained by the following formula: ； Among them, Fau is the comprehensive danger probability of the u-th person; fu1 is the first danger probability of the sub-area where the u-th person is currently located; fu2 is the second danger probability of the u-th person.

8. The monitoring method according to claim 1, characterized in that: The S6 includes: A first abnormality is obtained according to the AI ​​analysis result, and a first alarm is issued according to the first abnormality; If the first processing result indicates that the person's health is abnormal, a second alarm is issued; If any one of the first danger probability, the second danger probability or the comprehensive danger probability exceeds the corresponding probability threshold, an early warning is issued for personnel safety.

9. A system for implementing the operation monitoring method based on a smart helmet according to claim 1, characterized in that: The system comprises: Configuration module, used to obtain configuration information of smart helmets and operators; The acquisition module is used to divide the work site into multiple sub-areas, collect work site information through the smart safety helmet, and the work site information includes images, videos and field data. The GPU unit performs AI analysis on the field images and videos to obtain AI analysis results; A first transmission module is used to obtain a second score of the sub-area of ​​the work site according to the AI ​​analysis result; and upload the work site information and the AI ​​analysis result to the edge processing device according to the first score, the second score and the data score of the sub-area; A first processing module, configured to perform a first processing on the work site information through an edge processing device to obtain a first processing result; A second processing module is used to upload the first processing result to the cloud platform, and process it through the cloud platform to obtain a second processing result; the second processing result includes: the first danger probability of the sub-area, the second danger probability of the personnel, and the comprehensive danger probability; An early warning module, used for issuing an early warning or an alarm according to the AI ​​analysis result, the first processing result and / or the second processing result; The first transmission module comprises: Each sub-area is scored based on the complexity and dangerousness of the work content, and a first score is obtained; Obtain a second score for each sub-area work site based on the AI ​​analysis results; The second rating is: ； Wherein, Pj2 is the current second score of the j-th sub-region; The risk factor score obtained for the i-th smart helmet in the j-th sub-area based on the current AI analysis results; According to the first score, the second score and the data score, a comprehensive data score is obtained; The comprehensive rating is: ； ； Pzkv is the comprehensive score of the k-th helmet's v-th group data; Pskv is the score of the k-th helmet's v-th group data; Pj1' is the first score of the k-th helmet's current j-th sub-region; Pj2' is the second score of the k-th helmet's current j-th sub-region; Pskv, Pj1', Pj2' are all greater than 0; x1 and x2 are both the first weight coefficients; Sskv is the real-time score of the k-th helmet's v-th group data; Bskv is the score of the change between the k-th helmet's v-th group data and the previous sampling data; According to the comprehensive data score, the transmission order of the data collected by the smart helmet is determined; the higher the score, the higher the priority of transmission; According to the transmission sequence, the data collected by the smart helmet is uploaded to the edge processing device.

Images