A health state monitoring method for employees on site attendance

Abstract

This invention discloses a method for monitoring employee health status through on-site attendance tracking, comprising: when employees clock in using the on-site attendance system, the air filter automatically adjusts its filtration parameters to reduce dust and other suspended particles; the attendance system acquires real-time heart rate, body temperature, and mental state data of employees through built-in video sensors and monitoring equipment; the body temperature data is analyzed during clocking in to detect any abnormal health conditions, and a decision tree algorithm is used to analyze the heart rate data after cleaning to assess the employee's actual work pressure; the attendance system simultaneously acquires data on the day's work content and work environment at the construction site, including noise levels and air quality; the acquired environmental data is fused with the previously analyzed body temperature and heart rate data, and a decision tree algorithm is used for a comprehensive health status assessment.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for monitoring the health status of employees who clock in and out at construction sites. Background Technology

[0002] With the rapid development of the global construction industry, numerous construction sites are scattered across the globe. The safety and health of construction workers have always been a focus of attention due to the variable environment and numerous unknown safety risks present at construction sites. Traditional attendance systems primarily manage employee attendance and ensure accurate wage payments, but this method only addresses time management and cannot assess the actual health status of employees or the quality of the working environment. In recent years, the accumulation of pollutants such as dust and particulate matter at construction sites has led to increasingly deteriorating air quality, increasing health risks for workers. Furthermore, high-frequency noise, mechanical vibrations, and other environmental factors at construction sites can also affect the acquisition of real-time physiological health data, resulting in inaccurate data. Although some monitoring equipment and tools exist for assessing the environmental conditions of construction sites, most systems lack real-time analysis and feedback capabilities, making it difficult to provide workers with real-time health warnings and protective recommendations. Simultaneously, existing health monitoring equipment often focuses on single physiological parameters, such as heart rate and body temperature, neglecting the complex relationship between environmental factors and physiological responses. This singular data monitoring approach cannot provide construction site managers with comprehensive health assessments and adjustment strategies. Furthermore, the diverse sources and formats of data at construction sites make data cleaning and integration a major challenge. Without a unified data processing method, individual data sources struggle to provide accurate information for decision-makers. Therefore, current technology lacks an intelligent check-in method that comprehensively considers the construction site environment, employee physical conditions, and work content, making it difficult for site managers to make accurate decisions and threatening the health and safety of workers. Summary of the Invention

[0003] This invention provides a method for monitoring the health status of employees who clock in and out at construction sites, mainly including: When employees clock in at the construction site using the system, the air filter automatically adjusts its parameters to reduce dust and other suspended particles. The system uses built-in video sensors and monitoring equipment to acquire real-time heart rate, body temperature, and mental state data. A built-in acoustic module listens to the surrounding environment; if high-frequency noise is detected, the system automatically applies a Fast Fourier Transform algorithm to analyze the sound frequency, filters out the noise, and cleans all acquired physiological and environmental parameter data, removing erroneous values ​​and standardizing the format. Body temperature data is analyzed during clocking in to detect any abnormal health conditions, and a decision tree algorithm is used to analyze the cleaned data. Heart rate data is used to assess employees' actual work stress; the attendance system simultaneously acquires daily work content and work environment data from the construction site, including noise levels and air quality; the acquired environmental data is fused with previously analyzed body temperature and heart rate data, and a decision tree algorithm is used for comprehensive health status assessment; based on the results of the comprehensive health status assessment, it is analyzed whether the employee's work tasks for the day match their current health status and whether preventative measures are needed for work-related illnesses or accidents; the fused health data is matched with real-time work content and environmental data from the construction site to generate a list of preventative measures and automatically display personalized preventative suggestions to employees.

[0004] In some embodiments, when employees clock in at the construction site using the clock-in system, the air filter automatically adjusts its filtration parameters to reduce dust and other suspended particles, including: If the employee attendance system records attendance, it retrieves the employee's attendance information and attendance time via API and sends a start request to the air filter via API. Upon receiving the start request from the API, the air filter activates its built-in cleaning mode. The attendance system uses internal sensors to acquire real-time data on the current dust and particulate matter concentration at the construction site. Based on the acquired real-time dust and particulate matter concentration data, it adjusts the air filter's cleaning mode, filtration level, and filter running time in real time. After the air filter finishes running, it uses the attendance system's internal sensors to measure the treated dust and particulate matter concentration. Based on the employee's attendance time and the real-time measured dust and particulate matter concentration data, a dust and particulate matter concentration prediction model is established using a support vector machine algorithm to predict the dust and particulate matter concentration at different attendance times. Based on the predicted dust and particulate matter concentration at different attendance times, the automatic control system adjusts the air filter's preset start mode.

[0005] In some embodiments, the attendance system acquires real-time heart rate, body temperature, and mental state data of employees through built-in video sensors and monitoring devices, including: The system acquires employee facial images using a built-in video sensor, extracts facial regions from the images using a Haar cascade classifier, and locates key markers on the facial images using a Dlib 68-point model. A feature vector is calculated based on the distance and angle between the key markers. The extracted facial features are compared with a known employee facial database to determine the facial data with the highest matching degree, and the employee's identity is returned. Real-time body temperature and heart rate data are collected using monitoring equipment. Video footage of employee activities is acquired using the built-in video sensor, and image frames are extracted from the video. Based on the extracted activity image frames, employee physical activity data, including range and frequency of movement, is obtained. The employee's mental state is assessed based on the physical activity data. The collected real-time employee heart rate, body temperature, and mental state data are cleaned to remove outliers and noise. The employee heart rate, body temperature, and mental state data are merged, and a data encryption module is activated to randomly generate an encryption key to encrypt the merged data. The merged data is then transmitted to the data storage system. After successful storage, a timestamp module is triggered to record the exact time of data storage.

[0006] In some embodiments, the attendance system monitors the surrounding environment through a built-in acoustic module. If high-frequency noise is detected, the system automatically applies a Fast Fourier Transform algorithm to analyze the sound wave frequency, filters out the noise, and performs data cleaning on all acquired physiological and environmental parameter data, removing erroneous values ​​and standardizing the format, including: The system acquires ambient sound data through its built-in acoustic module. A Fourier transform algorithm is used to convert the time-domain signal into the frequency-domain data. For the transformed frequency-domain data, a predetermined threshold is set. If the intensity of a frequency exceeds this threshold, it is considered noise. For frequencies identified as noise, a noise shielding filter is triggered. After noise shielding, employee physiological and environmental parameter data are acquired. Physiological parameters include heart rate and body temperature, while environmental parameters include ambient temperature, humidity, sound frequency, and dust and particulate matter concentration. Simultaneously, the sound wave frequency characteristics identified by the Fourier transform algorithm are used to annotate the physiological and environmental data. Once annotation is complete, the physiological and environmental data are cleaned, erroneous values ​​are removed, and the data format is standardized.

[0007] In some embodiments, the process of clocking in involves analyzing body temperature data to detect any abnormal health conditions, and using a decision tree algorithm to analyze cleaned heart rate data to assess the employee's actual work stress, including: During check-in, the system acquires employee IDs and receives employee body temperature data via API. Using preset temperature thresholds, it analyzes the data to determine if employees have fever or hypothermia. Employees with abnormal temperatures are flagged and added to a special observation database. Using these employee IDs, the system retrieves corresponding heart rate data via API. The heart rate data is cleaned using median denoising to remove outliers and noise. Based on employee heart rate, work hours, rest time, work content, and work environment data, a decision tree algorithm is used to classify employee work stress and determine their current stress level. The work stress classification results associated with the employee IDs are stored in the analysis results database. If the work stress assessment indicates a high-pressure state, the employee ID is marked as requiring further attention and observation. An intersection operation is performed on the special observation database and the analysis results database to identify employees with both abnormal body temperature and work stress. The system also includes determining the employee's work stress level using a decision tree algorithm based on their heart rate, work hours, rest time, work content, and work environment data.

[0008] The process involves using a decision tree algorithm to determine an employee's work stress level based on data such as heart rate, working hours, rest time, job content, and work environment. Specifically, this includes: Based on employee heart rate data, the average, maximum, minimum, and standard deviation of heart rate are obtained. Daily work hours and rest periods for employees are also obtained. Based on employee job content, job type and workload are determined. Based on employee work environment, environmental conditions and noise levels are collected. Feature selection is performed using the Pearson correlation coefficient method to identify attributes that significantly influence work stress classification, and these are then filtered to obtain the most relevant attribute set. The dataset is divided into training and test sets, and a decision tree algorithm is used to construct a decision tree model based on the attribute labels of the training set. The constructed decision tree model is then used to predict the work stress level of employees.

[0009] In some embodiments, the attendance system synchronously acquires the work content and working environment data of the construction site for the day, including noise level and air quality, including: The system synchronously acquires daily work content data from the construction site via data interfaces and automatically sorts tasks according to priority. It also synchronously acquires noise level and air quality data from the construction site; if noise levels exceed preset thresholds, the corresponding alarm system is automatically triggered; if air quality data falls below preset levels, environmental purification equipment is automatically activated. Furthermore, it synchronously acquires the number of workers on the construction site and automatically assigns tasks based on work content and number of workers. The system acquires the location of the construction site and nearby sites and assesses the possibility of resource sharing based on geographical location data. It synchronously acquires the status of all working equipment via sensors, automatically marking non-working equipment as standby and monitoring its power consumption. Finally, it synchronously acquires the completion status and duration of work tasks and automatically adjusts future task priorities based on work duration and content.

[0010] In some embodiments, the step of fusing the acquired environmental data with previously analyzed body temperature and heart rate data, and using a decision tree algorithm to perform a comprehensive health status assessment, includes: Work environment data is acquired using environmental sensors; employee body temperature and heart rate data are acquired using wearable devices; missing values ​​in environmental, body temperature, and heart rate data are imputed using mean imputation; Z-score standardization is used to standardize the four imputed data sets; the most relevant features to health status are selected from the standardized data based on the correlation matrix; the four data sets are integrated into a unified dataset using the selected features; a decision tree algorithm is used to train the model on the integrated dataset; the newly acquired data is then used to assess health status based on the trained decision tree model; if the output is below a predetermined threshold, it is marked as requiring further analysis; Z-score standardization is used to standardize the marked data; the decision tree model is retrained using the newly processed data; and the newly acquired data is then used to assess health status based on the retrained decision tree model.

[0011] In some embodiments, analyzing whether an employee's work tasks for the day match their current health status and whether preventative measures against work-related illnesses or accidents are necessary based on the results of a comprehensive health status assessment includes: The system extracts daily environmental data, body temperature, and heart rate data from the database. Using a trained decision tree model, it performs a preliminary analysis of employee health status data to obtain an assessment. It then acquires the employee's daily work task data and determines whether the task matches their current health status. If the task does not match, it is reassigned. Based on the environmental, body temperature, and heart rate data, and the work task data, it determines whether the employee's task is high-risk; if so, the employee is in a high-risk state. It also determines whether the workload exceeds the employee's physical or psychological capacity; if so, the employee is in a high-risk state. Finally, it assesses whether the employee's work environment contains safety hazards or harmful substances; if so, the employee is in a high-risk state. Finally, it checks whether the employee's body temperature and heart rate are within acceptable limits; if not, the employee is in a high-risk state. If a high-risk employee is identified, a high-risk warning module is triggered, and the employee's information is added to a monitoring list to prevent related illnesses or accidents.

[0012] In some embodiments, the process of matching the fused health data with real-time work content and environmental data at the construction site to generate a list of preventative measures and automatically display personalized preventative recommendations to employees includes: The system acquires fused health data from multiple databases and sensors via API; it obtains real-time work content text from construction site logs and extracts key features using the TF-IDF algorithm; based on the extracted features, the work content is classified into predetermined categories, and one or more labels are assigned to each task, describing the nature and requirements of the task; the classification and labeling results are converted into structured work content data in database format; environmental data from the construction site, including temperature, humidity, and pollutant concentration, is acquired through environmental monitoring equipment; the fused health data, work content data, and environmental data are preprocessed to unify the data format, and the preprocessed data is matched and merged to form a comprehensive dataset; a simple linear regression algorithm is used to analyze the comprehensive dataset, and by judging the implementation and effectiveness of preventive measures, data features related to preventive measures are obtained; based on the obtained data features, a list of preventive measures is developed; based on the list of preventive measures and the real-time work content of construction site employees, personalized preventive recommendations are developed for each employee; and the personalized preventive recommendations for each employee are sent to each employee's mobile device.

[0013] The present invention has the following beneficial effects: This invention discloses a method for monitoring the health status of employees during on-site attendance tracking. It acquires real-time heart rate, body temperature, and mental state data of employees through built-in video sensors and monitoring equipment, and uses a built-in acoustic module to monitor the surrounding environment, analyze it in real time, and filter out high-frequency noise.

[0014] During the check-in process, the system analyzes body temperature data to detect any abnormal health conditions and uses a decision tree algorithm to analyze cleaned heart rate data to assess the employee's actual work stress. Simultaneously, the system acquires data on the day's work content and work environment at the construction site.

[0015] The acquired environmental data was fused with previously analyzed body temperature and heart rate data, and a decision tree algorithm was used to conduct a comprehensive health status assessment. Based on the assessment results, it was analyzed whether the employee's work tasks for the day matched their current health status and whether preventative measures against work-related illnesses or accidents were necessary.

[0016] Ultimately, the system matches the integrated health data with real-time work content and environmental data from the construction site to generate a list of preventative measures and automatically displays personalized preventative recommendations to employees. Through this integrated technology, employees can promptly understand their own health status and take appropriate measures, thereby improving work efficiency and preventing work-related health problems and accidents. Attached Figure Description

[0017] Figure 1 This is a flowchart of a method for monitoring employee health status through on-site attendance tracking, according to the present invention.

[0018] Figure 2 This is a schematic diagram of an employee health status monitoring method for on-site attendance tracking according to the present invention.

[0019] Figure 3 This is another schematic diagram of an employee health status monitoring method for on-site attendance tracking according to the present invention. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to the accompanying drawings. The embodiments described with reference to the drawings are exemplary and intended to explain the present invention, but should not be construed as limiting the present invention.

[0021] This embodiment of a method for monitoring employee health status through construction site attendance tracking may specifically include: In step S101, when employees clock in at the construction site's clock-in system, the air filter automatically adjusts its filtration parameters to reduce dust and other suspended particles.

[0022] If the employee attendance system records an attendance event, it retrieves the employee's attendance information and attendance time via API and sends a start request to the air filter via API. Upon receiving the start request from the API, the air filter activates its built-in cleaning mode. The attendance system uses internal sensors to acquire real-time data on the current dust and particulate matter concentration at the construction site. Based on this real-time data, it adjusts the air filter's cleaning mode, filtration level, and operating time. After the air filter finishes operating, its internal sensors measure the treated dust and particulate matter concentration. Based on the employee's attendance time and the real-time dust and particulate matter concentration data, a dust and particulate matter concentration prediction model is established using a support vector machine algorithm to predict the dust and particulate matter concentration at different attendance times. Based on the predicted dust and particulate matter concentration at different attendance times, the automated control system adjusts the air filter's preset start mode. For example, if employee A attends at 8:00 AM, the system records the attendance event and retrieves employee A's attendance information and attendance time (8:00 AM) via API. The attendance system sends a start request to the air filter via API. Upon receiving the start request, the air filter activates its built-in cleaning mode. Based on real-time sensor data of dust and particulate matter concentrations at the construction site (dust concentration 1000 μg / m³, particulate matter 50 μg / m³), the attendance system adjusts the air filter's cleaning mode, filtration level, and runtime in real time. After the air filter finishes its operation, the system's internal sensors measure the post-treatment dust and particulate matter concentrations. The post-treatment dust concentration is 500 μg / m³, and the particulate matter concentration is 20 μg / m³. Based on employee A's attendance time and the real-time dust and particulate matter concentration data, a support vector machine algorithm is used to build a dust and particulate matter concentration prediction model. The model predicts that at 8:00 AM, employee A's dust concentration will be 600 μg / m³, and the particulate matter concentration will be 30 μg / m³. Based on the predicted dust and suspended particulate concentrations at different check-in times, the automated control system can adjust the preset start-up mode of the air filter, such as increasing the running time of the filter cleaning mode and improving the filtration level when high dust and suspended particulate concentrations are predicted.

[0023] In step S102, the attendance system acquires real-time heart rate, body temperature, and mental state data of employees through built-in video sensors and monitoring equipment.

[0024] Employee facial images are acquired using a built-in video sensor. A Haar cascade classifier is used to extract facial regions from the detected images. A 68-point model from Dlib is used to locate key markers on the facial images. A feature vector is calculated based on the distance and angle between the key markers. The extracted facial features are compared with a known employee facial database to determine the facial data with the highest matching degree, and the employee's identity is returned. Real-time body temperature and heart rate data of employees are collected using monitoring equipment. Employee activity videos are acquired using the built-in video sensor, and image frames are extracted from the videos. Based on the extracted activity image frames, employee physical activity data, including the range and frequency of physical activity, is obtained. Employee mental state is assessed based on the physical activity data. The collected real-time employee heart rate, body temperature, and mental state data are cleaned to remove outliers or noise. The employee heart rate, body temperature, and mental state data are merged, and a data encryption module is activated to encrypt the merged data using a randomly generated encryption key. The merged data is then transmitted to the data storage system. After successful storage in the data storage system, a timestamp module is triggered to record the exact time of data storage. For example, the facial region feature vector extracted from the employee's facial image acquired by the built-in video sensor is [2,5,8,3]. When compared with facial data in the known employee facial database, the facial data with the highest matching degree is found, and its feature vector is [2,4,7,2]. Additionally, the collected real-time employee body temperature data is 35 degrees Celsius, and the real-time heart rate data is 80 beats per minute. Body activity data obtained from image frames extracted from employee activity videos shows that the employee's body activity range is 5 square meters, and the body activity frequency is 10 times per minute. Based on the body activity data, the employee's mental state is judged to be normal. After data cleaning to remove outliers or noise, the employee's heart rate is 78 beats per minute, body temperature is 33 degrees Celsius, and mental state remains normal. Finally, the employee's heart rate, body temperature, and mental state data are encrypted using a random encryption key generated by the data encryption module and transmitted to the data storage system. After the data is successfully stored in the data storage system, the timestamp module records the exact time of data storage as 10:30 AM on January 1, 2022.

[0025] In step S103, the attendance system monitors the surrounding environment through its built-in acoustic module. If high-frequency noise is detected, the system automatically applies the Fast Fourier Transform algorithm to analyze the sound wave frequency, and blocks the noise. It also cleans all the acquired physiological and environmental parameter data, deletes erroneous values, and standardizes the format.

[0026] The system acquires ambient sound data through its built-in acoustic module. A Fourier transform algorithm is used to convert the time-domain signal into the frequency-domain data. For the transformed frequency-domain data, a pre-defined threshold is set. If the intensity of a frequency exceeds this threshold, it is considered noise. For frequencies identified as noise, a noise-masking filter is triggered for noise masking. After noise masking, employee physiological and environmental parameter data are acquired. Physiological parameters include heart rate and body temperature, while environmental parameters include ambient temperature, humidity, sound frequency, and dust and particulate matter concentration. Simultaneously, the physiological and environmental data are labeled using sound wave frequency features identified by the Fourier transform algorithm. Once labeling is complete, the physiological and environmental data undergoes data cleaning, removing erroneous values ​​and standardizing the data format. For example, the acoustic module of the attendance system acquires ambient sound data and applies a Fourier transform algorithm to convert it into frequency domain data [100Hz, 200Hz, 300Hz, 400Hz, 500Hz], setting a predetermined threshold of 300Hz. Based on the threshold, it can be determined which frequencies are considered noise. Therefore, the intensities of 400Hz and 500Hz both exceed the threshold of 30, so 400Hz and 500Hz can be determined as noise frequencies. Next, a noise shielding filter is triggered to shield the noise. The frequency domain data after shielding is [100Hz, 200Hz, 300Hz]. Then, the employee's physiological and environmental parameter data can be obtained: the employee's heart rate is 80 bpm, and the body temperature is 35 degrees Celsius. The environmental parameter data is: temperature 25 degrees Celsius, humidity 40%, sound frequency 300Hz, dust concentration 0.2 mg / m³, and suspended particulate concentration 10 ppm. Finally, the physiological and environmental data are labeled using the sound wave frequency features identified by the Fourier transform algorithm. For example, heart rate, body temperature and sound frequency are labeled with their corresponding frequencies, such as heart rate of 80 bpm - frequency of 300 Hz and body temperature of 35 degrees Celsius - frequency of 200 Hz. After labeling, the physiological and environmental data are cleaned, erroneous values ​​are deleted and the data format of physiological and environmental data is standardized.

[0027] Step S104: Analyze body temperature data during check-in to detect any abnormal health conditions, and use a decision tree algorithm to analyze the cleaned heart rate data to assess the employee's actual work pressure.

[0028] During check-in, the system retrieves employee IDs and receives employee body temperature data via API. Using preset temperature thresholds, the system analyzes the temperature data to determine if the employee has a fever or hypothermia. Employees with abnormal temperatures are flagged and added to a special observation database. Using these employee IDs, the system retrieves corresponding heart rate data via API. The heart rate data is cleaned using median denoising to remove outliers and noise. Based on employee heart rate, work hours, rest time, job content, and work environment data, a decision tree algorithm is used to classify employee work stress and determine their current stress level. The work stress classification results associated with the employee IDs are stored in the analysis results database. If the work stress assessment indicates a high-pressure state, the employee ID is marked as requiring further attention and observation. An intersection operation is performed on the special observation database and the analysis results database to identify employees with both abnormal body temperature and abnormal work stress. For example, during check-in, employee ID 1001 is retrieved, and their body temperature data (35℃) is received via API. Based on preset body temperature thresholds, a body temperature below 36℃ or above 35℃ is considered an abnormal temperature. Employee 1001 has a normal body temperature, with no fever or hypothermia. For heart rate data, based on employee information for employee 1001 in the special observation database, the corresponding heart rate data is obtained via API. The obtained heart rate data is [70, 80, 75, 100, 85]. Using median denoising, the median of the heart rate data is calculated to be 80. If a value in the heart rate data differs from the median by more than a preset threshold of 10, it is considered an outlier or noise. The heart rate value of 100 exceeds the threshold of the median of 80 plus 10, therefore it is considered an outlier and removed. Based on employee 1001's heart rate data [70, 80, 75, 85], working hours of 8 hours, rest time of 1 hour, job content of office work, and working environment of a comfortable office environment, a decision tree algorithm is used to classify work stress. Based on the algorithm model, employee 1001's work stress was classified as normal. Employee ID 1001 and the work stress classification result were stored in the analysis results database. An intersection operation was performed on the special observation database and the analysis results database, revealing no employees with both abnormal body temperature and abnormal work stress.

[0029] Based on data on employee heart rate, working hours, rest time, job content, and work environment, a decision tree algorithm is used to determine the employee's work stress level.

[0030] Based on employee heart rate data, the average, maximum, minimum, and standard deviation of heart rate are obtained. Daily work hours and rest periods for employees are also obtained. Based on employee job content, job type and workload are determined. Based on employee work environment, environmental conditions and noise levels are collected. Feature selection is performed using the Pearson correlation coefficient method to identify attributes that significantly influence work stress classification, and these are then filtered to obtain the most relevant attribute set. The dataset is divided into training and test sets, and a decision tree algorithm is used to construct a decision tree model based on the attribute labels of the training set. The constructed decision tree model is then used to predict the work stress level of employees. For example, we have data for 5 employees: (Employee, Heart Rate, Working Hours, Rest Time, Job Type, Work Intensity, Working Environment Conditions, Noise Level, Stress Level). (A, 70, 8, 1, Cleaning the Site, Medium, Good, 50, Low); (B, 85, 10, 0.5, Handling Materials, High, Average, 60, Medium); (C, 78, 9, 1.5, Repairing Equipment, Low, Good, 45, Low); (D, 82, 11, 0.5, Handling Materials, High, Poor, 70, High); (E, 75, 7, 2, Cleaning the Site, Low, Good, 48, Low). Therefore, the average heart rate data is 78, the maximum is 85, the minimum is 70, and the standard deviation is approximately 5.25. Based on the table, we obtain the working hours and rest time, encode the job type and work intensity as numbers, encode the working environment conditions as Good=1, Average=2, Poor=3, and use decibel values ​​for noise levels. Feature selection used Pearson correlation coefficients, which revealed that "working hours" and "noise level" had the highest correlation with "stress level," and these two features were selected. A, B, and C were used as the training set, and D and E as the test set. Using the training set data, a simple decision tree model was constructed to predict "stress level" based on "working hours" and "noise level." The decision tree model was used to predict stress levels for data in D and E, and then compared with actual stress levels to evaluate the model's accuracy. Based on the test results, the model's accuracy was determined, and whether further adjustments to the model parameters were needed were assessed.

[0031] Step S105: The attendance system synchronously acquires the work content and working environment data of the construction site for the day, including noise level and air quality.

[0032] The system synchronously acquires daily work data from the construction site via a data interface and automatically sorts tasks according to priority. It also synchronously acquires noise level and air quality data; if noise levels exceed preset thresholds, the corresponding alarm system is automatically triggered; if air quality falls below preset levels, environmental purification equipment is automatically activated. The system synchronously acquires the number of workers on the site and automatically assigns tasks based on work content and number of workers. It acquires the location of the construction site and nearby sites and assesses resource sharing possibilities based on geographical location data. Sensors synchronously acquire the status of all work equipment, automatically marking inactive equipment as standby and monitoring battery levels. The system synchronously acquires the completion status and duration of work tasks and automatically adjusts future task priorities based on duration and content. For example, at 8:00 AM every day, the system automatically acquires daily work data from the construction site. If there are three tasks that day: site cleaning, equipment repair, and material handling, site cleaning is prioritized first, equipment repair second, and material handling third. The noise level at the construction site is 80 decibels, exceeding the preset threshold of 70 decibels, automatically triggering the corresponding alarm system. The air quality at the construction site has a PM5 concentration of 50 micrograms per cubic meter, below the preset level of 60 micrograms per cubic meter, automatically activating the environmental purification equipment. There are 10 workers on the construction site. Cleaning the site requires 3 workers, equipment repair requires 5 workers, and material handling requires 2 workers. The system will automatically assign workers to the corresponding tasks. Construction site A is located in zone H, with construction sites B and C nearby. Based on geographical location data, it was determined that construction sites A and B can share some equipment and manpower, while construction site C is too far away, making resource sharing less likely. Sensors detected that device D is not in operation; the system will automatically mark device D as standby and check its battery level. The site cleaning task was estimated to take 2 hours to complete, but the actual completion time was 3 hours. The system will automatically adjust the priority of future site cleaning tasks to ensure timely completion.

[0033] Step S106: The acquired environmental data is fused with the previously analyzed body temperature and heart rate data, and a decision tree algorithm is used to perform a comprehensive health status assessment.

[0034] Environmental sensors were used to acquire work environment data. Wearable devices worn by employees were used to acquire employee body temperature and heart rate data. Mean imputation was used to complete missing values ​​in the environmental, body temperature, and heart rate data. Z-score normalization was used to standardize the four groups of data after imputation. Features most relevant to health status were selected from the standardized data based on the correlation matrix. Using the selected features, the four groups of data were integrated into a unified dataset. A decision tree algorithm was used to train a model on the integrated dataset. Based on the trained decision tree model, the newly acquired data was used to assess health status. If the output result was below a predetermined threshold, it was marked as requiring further analysis. Z-score normalization was used to standardize the marked data; the decision tree model was retrained using the newly processed data; and the newly acquired data was used to assess health status based on the retrained decision tree model. For example, environmental sensors were used to acquire daily data on temperature, humidity, sound frequency, dust, and suspended particulate concentration. In addition, employees wore wearable devices to acquire daily body temperature and heart rate data. Data was collected for one month, but the number of tasks for one day was missing. The average number of tasks for other days was calculated and used as the missing task count. Next, the four completed datasets were standardized. Then, based on the correlation matrix, the features most relevant to health status were selected. Body temperature and heart rate were found to have the highest correlation. These two features were selected and integrated with the work content and environmental data into a unified dataset. A decision tree algorithm was used to train the model on the integrated dataset, using the training set data to train the decision tree, resulting in a model with high accuracy. Next, the latest data for one employee was obtained: temperature 25℃, humidity 60%, body temperature 35℃, and heart rate 80 beats / minute. According to the decision tree model, their health status was assessed as normal. Employees with a health status score below 80 were marked as requiring further analysis. These marked data were standardized, and the decision tree model was retrained. Finally, the newly acquired data was used to assess health status based on the retrained decision tree model.

[0035] Step S107: Based on the results of the comprehensive health status assessment, analyze whether the employee's work tasks for the day match their current health status and whether it is necessary to take preventive measures against work-related illnesses or accidents.

[0036] The system extracts daily environmental data, body temperature, and heart rate data from the database. Using a trained decision tree model, it performs a preliminary analysis of employee health status data to obtain an assessment. It then acquires the employee's daily work task data and determines whether the task matches their current health status. If not, the task is reassigned. Based on the environmental, body temperature, and heart rate data, the system assesses whether the employee's task is high-risk; if so, the employee is considered high-risk. It also determines whether the workload exceeds the employee's physical or psychological capacity; if so, the employee is considered high-risk. Furthermore, it checks the work environment for potential safety hazards or harmful substances; if present, the employee is considered high-risk. Finally, it checks whether the employee's body temperature and heart rate are within safe limits; if not, the employee is considered high-risk. If a high-risk employee is identified, a high-risk warning module is triggered, and the employee's information is added to a monitoring list. Preventative measures are then implemented to address any related illnesses or accidents affecting employees in a high-risk state. For example, consider the environmental data for employee Xiao Zhang on a certain day: temperature 25℃, humidity 60%, CO2 concentration 800ppm. Xiao Zhang's body temperature and heart rate data are: body temperature 37℃, heart rate 80 beats / minute. Xiao Zhang's task that day was to move materials. First, a pre-trained decision tree model is used to conduct a preliminary analysis of Xiao Zhang's health status. According to the model, Xiao Zhang's health status is assessed as healthy. Next, based on Xiao Zhang's health status and work task data, it is determined whether the work task matches his current health status. Since Xiao Zhang's health status is healthy, it matches the task of moving materials. Then, based on the environmental data, body temperature and heart rate data, and work task data, it is determined whether the work task type is a high-risk task. Moving materials is defined as a low-risk task, therefore Xiao Zhang's work task type is not a high-risk task. Next, it is determined whether the intensity of the work task exceeds his physical or psychological tolerance. The intensity of moving materials is low, and Xiao Zhang's physical and psychological tolerance is within the normal range, therefore the intensity of the work task does not exceed his tolerance. Finally, it is determined whether there are any safety hazards or harmful substances in the work environment. The surrounding environment for handling materials is safe and free of harmful substances; therefore, the work environment does not present any safety hazards or harmful substances. Next, the body temperature and heart rate are assessed. The standard range for body temperature and heart rate is 35℃-35℃ and 60-100 beats / minute, respectively. Since Xiao Zhang's body temperature and heart rate are within the standard range, they are considered acceptable. Finally, based on the above criteria, Xiao Zhang's work task is not a high-risk task, the intensity of the task does not exceed his tolerance, the work environment does not present any safety hazards or harmful substances, and his body temperature and heart rate are within the standard range; therefore, Xiao Zhang is not in a high-risk state.If a high-risk employee is identified, such as employee Xiao Li whose environmental data, body temperature and heart rate data, and work tasks are the same as Xiao Zhang's, but whose body temperature is 38℃, then based on the above criteria, Xiao Li's body temperature is below standard, placing him in a high-risk state. The system will trigger the high-risk warning module and add Xiao Li's information to the monitoring list to prevent potential related illnesses or accidents.

[0037] Step S108: The integrated health data is matched with the real-time work content and environmental data of the construction site to generate a list of preventive measures and automatically display personalized preventive suggestions to employees.

[0038] The system acquires fused health data from multiple different databases and sensors via API. Real-time work content text from construction site logs is obtained, and key features are extracted using the TF-IDF algorithm. Based on these features, work content is categorized into predetermined classes, and one or more labels are assigned to each task, describing its nature and requirements. The classification and labeling results are then transformed into structured work content data in a database format. Environmental data from the construction site, including temperature, humidity, and pollutant concentration, is acquired through environmental monitoring equipment. The fused health data, work content data, and environmental data are preprocessed to standardize the data format, and then matched and merged to form a comprehensive dataset. A simple linear regression algorithm is used to analyze the comprehensive dataset, determining the implementation and effectiveness of preventative measures and obtaining data features related to these measures. Based on these data features, a list of preventative measures is developed. Personalized preventative recommendations are then developed for each employee based on the preventative measures list and the real-time work content of construction site workers. These personalized recommendations are sent to each employee's mobile device. For example, fused health data is acquired via API from 20 different databases and 50 sensors. One of the databases records each person's height and weight information, allowing for the calculation of average height and weight. The database contains height data for 5 individuals: [170, 175, 180, 165, 172] cm, and weight data: [60, 65, 70, 55, 58] kg. The average height can be calculated as (170+175+180+165+172) / 5 = 172.4 cm, and the average weight as (60+65+70+55+58) / 5 = 61.6 kg. Real-time work content text from the construction site is obtained through site logs, and the TF-IDF algorithm is used to extract key information and features from the text. One work content text is "moving heavy objects to the construction site," and its TF-IDF value can be calculated to extract key information. The keywords in the text are "moving" and "construction site," and their TF-IDF values ​​can be calculated. "Moving" appears 3 times in the text and 10 times in the entire text set; "construction site" appears 2 times in the text and 20 times in the entire text set. The TF-IDF value for "handling" is 3 / 10 = 0.3, and the TF-IDF value for "construction site" is 2 / 20 = 0.1. Based on the extracted features, the work content is classified into predetermined categories, and one or more labels are assigned to each task, describing the nature and requirements of the task. Work content can be divided into three categories: handling, cleaning, and installation. One work content is classified as handling and labeled as "heavy objects" and "manual labor." The classification and labeling results are converted into structured work content data in a database format. An environmental monitoring device recorded the site temperature as 25℃, humidity as 60%, and PM5 concentration as 50 μg / m³.The merged health data, job content data, and environmental data were preprocessed to unify the data format. Health data included height and weight, job content data included classification and labels, and environmental data included temperature, humidity, and pollutants. These were integrated into a single dataset, with the data format unified as [height, weight, classification, label, temperature, humidity, PM5]. The preprocessed data were matched and merged to form a comprehensive dataset. One data point in the comprehensive dataset is [174, 66,"handling", "heavy objects, manual labor", 25, 60, 50], indicating that an employee's height is 174cm, weight is 66kg, job content is handling heavy objects, labels are heavy objects and manual labor, site temperature is 25℃, humidity is 60%, and PM5 concentration is 50μg / m³. A simple linear regression algorithm was used to analyze the comprehensive dataset to obtain data features related to preventive measures. Using simple linear regression to analyze the comprehensive dataset can yield data features related to preventive measures, such as the influence of weight on height and the influence of temperature on job content. Based on the impact of weight on height, a list of preventative measures can be developed, such as controlling employees' weight within a healthy range to reduce physical strain. Based on this list and each employee's job duties, personalized preventative recommendations can be created for each employee, such as advising employees lifting heavy objects to protect their lower backs and to manage work intensity appropriately. These personalized recommendations can then be sent directly to each employee's mobile device so they can view the recommendations and take appropriate preventative measures at any time.

[0039] Obviously, those skilled in the art can make various modifications and variations to the embodiments of this application without departing from the spirit and scope of the embodiments of this application. Therefore, if these modifications and variations to the embodiments of this application fall within the scope of the claims of this application and their equivalents, this application also intends to include these modifications and variations.

[0040] The above descriptions are merely some embodiments of the present invention. Those skilled in the art can make various modifications and improvements without departing from the inventive concept of the present invention, and these all fall within the scope of protection of the present invention.

Claims

1. A method for monitoring the health status of employees who clock in and out at construction sites, characterized in that, The method includes: When employees clock in at the construction site using the system, the air filter automatically adjusts its parameters to reduce dust and other suspended particles. The system uses built-in video sensors and monitoring equipment to acquire real-time heart rate, body temperature, and mental state data. A built-in acoustic module listens to the surrounding environment; if high-frequency noise is detected, the system automatically applies a Fast Fourier Transform algorithm to analyze the sound frequency, filters out the noise, and cleans all acquired physiological and environmental parameter data, removing erroneous values ​​and standardizing the format. Body temperature data is analyzed during clocking in to detect any abnormal health conditions, and a decision tree algorithm is used to analyze the cleaned data. Heart rate data is used to assess employees' actual work stress; the attendance system simultaneously acquires daily work content and work environment data from the construction site, including noise levels and air quality; the acquired environmental data is fused with previously analyzed body temperature and heart rate data, and a decision tree algorithm is used for comprehensive health status assessment; based on the results of the comprehensive health status assessment, it is analyzed whether the employee's work tasks for the day match their current health status and whether preventative measures are needed for work-related illnesses or accidents; the fused health data is matched with real-time work content and environmental data from the construction site to generate a list of preventative measures and automatically display personalized preventative suggestions to employees.

2. The method according to claim 1, wherein, When employees clock in / out using the construction site's attendance system, the air filter automatically adjusts its filtration parameters to reduce dust and other suspended particles, including: Employee attendance information is obtained via API, and a start request is sent to the air filter via API; the air filter starts its built-in cleaning mode; the attendance system obtains the current dust and suspended particulate concentration data of the construction site; the air filter's filter cleaning mode, filtration level, and running time are adjusted in real time; after the process is completed, the dust and suspended particulate concentration is measured; a dust and suspended particulate concentration prediction model is established, and the preset start mode of the air filter is adjusted.

3. The method according to claim 1, wherein, The attendance tracking system acquires real-time heart rate, body temperature, and mental state data of employees through built-in video sensors and monitoring devices, including: The system uses video sensors to acquire employee facial images and extract facial features; it uses Dlib's 68-point model to locate facial markers and identify employees; it monitors employee body temperature and heart rate; it acquires employee activity videos and extracts image frames to obtain employee physical activity data; it assesses employee mental state; it cleans the collected real-time employee heart rate, body temperature, and mental state data, merges the cleaned data, activates the encryption module, and transmits it to the data storage system, recording the data storage time.

4. The method according to claim 1, wherein, The attendance tracking system uses a built-in acoustic module to monitor the surrounding environment. If high-frequency noise is detected, the system automatically applies a Fast Fourier Transform algorithm to analyze the sound wave frequency, filters out the noise, and performs data cleaning on all acquired physiological and environmental parameter data, removing erroneous values ​​and standardizing the format, including: Acquire sound data through the acoustic module; convert the sound data to the frequency domain using Fourier transform; set a threshold to determine noise and trigger noise shielding; acquire employee physiological and environmental parameter data and label the sound wave frequency characteristics; and unify the physiological and environmental data formats after data cleaning.

5. The method according to claim 1, wherein, The check-in process involves analyzing body temperature data to detect any abnormal health conditions, and using a decision tree algorithm to analyze cleaned heart rate data to assess employees' actual work stress, including: The system acquires employee attendance tags and receives employee body temperature data via API. It uses preset temperature thresholds to determine if employees have fever or hypothermia. Employees with abnormal temperatures are flagged and added to a special observation database. Using these employee tags, the system retrieves corresponding heart rate data via API. The heart rate data is cleaned, and outliers are removed using median denoising. Based on employee heart rate, work hours, rest time, work content, and work environment data, a decision tree algorithm is used to classify employee work stress and determine their current stress level. The work stress classification results associated with the employee tags are stored in the analysis results database.

6. The method according to claim 1, wherein, The attendance system synchronously acquires the work content and working environment data of the construction site for the day, including noise level and air quality, including: The system synchronously acquires daily work data from the construction site via data interfaces; it also synchronously acquires noise level and air quality data from the construction site, triggering an alarm system if the noise level exceeds a preset threshold and activating environmental purification equipment if the air quality falls below a preset level; it synchronously acquires the number of workers on the construction site and automatically assigns work tasks; it acquires the location of the construction site and nearby construction sites to assess resource sharing; it synchronously acquires the status of all work equipment, marks non-working equipment as backup, and performs power level checks; and it synchronously acquires the completion status and duration of work tasks, automatically adjusting the priority of future tasks.

7. The method according to claim 1, wherein, The process of fusing the acquired environmental data with previously analyzed body temperature and heart rate data, and using a decision tree algorithm for comprehensive health status assessment, includes: Work environment data is acquired using environmental sensors; employee body temperature and heart rate data are acquired using wearable devices; missing values ​​in all data are imputed using mean imputation; the imputed data is standardized using Z-score standardization; the features most relevant to health status are selected based on the correlation matrix; all data are integrated into a unified dataset; the integrated dataset is trained using a decision tree algorithm; health status is assessed on new data; and data with output results below a predetermined threshold are marked as requiring further analysis.

8. The method according to claim 1, wherein, Based on the results of the comprehensive health status assessment, the analysis determines whether the employee's daily work tasks match their current health status and whether preventative measures are needed to prevent work-related illnesses or accidents, including: The system extracts daily environmental data, body temperature, and heart rate data from the database; analyzes employee health data using a decision tree model; obtains employee work task data and determines whether the tasks match their health status; if not, work tasks are reassigned; based on environmental data, body temperature and heart rate data, and work task data, it determines whether the employee's work task is a high-risk task, whether their body temperature and heart rate are within acceptable limits, and whether there are potential hazards in the work environment; if a high-risk employee is identified, an early warning module is triggered, and they are added to the monitoring list.

9. The method according to claim 1, wherein, The process involves matching the fused health data with real-time work content and environmental data at the construction site to generate a list of preventative measures and automatically display personalized preventative recommendations to employees, including: Health data is obtained through APIs; real-time work content is obtained through site logs, and text features are extracted using the TF-IDF algorithm; work content is classified into predetermined categories; environmental data is obtained through monitoring equipment; all data is preprocessed and matched to form a comprehensive dataset; the dataset is analyzed using a linear regression algorithm to obtain features related to preventive measures; a list of preventive measures is compiled and preventive recommendations are developed for employees; preventive recommendations are sent to employees' devices.

Images