Building construction safety risk assessment method and system

Abstract

The invention discloses a building construction safety risk assessment method and system, and relates to the technical field of artificial intelligence, and the method comprises the steps: constructing a neural network model, and recognizing an unsafe behavior mode of a worker; receiving a data stream from a construction site in real time, dynamically adjusting risk assessment parameters according to a current construction state and environment change, and predicting potential safety risks; according to the prediction result, generating corresponding prevention measure suggestions; key indexes and abnormal conditions are displayed through a monitoring interface, and a detailed analysis report is generated; according to the invention, through dynamic assessment and timely response to the safety risk of the construction site, the risk assessment parameters can be rapidly adjusted according to the environmental change, the potential risk can be predicted, and the prevention measures and suggestions can be provided, so that the occurrence probability of safety accidents is effectively reduced, and the construction efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a construction safety risk assessment method and system. Background Art

[0002] In recent years, the construction industry has generally adopted Internet of Things (IoT) technology and artificial intelligence (AI) to improve the level of construction safety management. With the support of IoT technology, various sensors are widely used to monitor the construction environment. In addition, intelligent algorithms based on machine learning have also been introduced to the construction site to identify potential safety hazards, such as workers who do not wear personal protective equipment (PPE), equipment failures, and non-compliant operating behaviors. Although these technologies have improved construction safety to a certain extent, they still face many challenges in actual application.

[0003] Although existing technologies have improved construction safety to a certain extent, there are still some shortcomings. First, most of the current intelligent safety monitoring systems can only make simple threshold judgments on environmental parameters, and lack the ability to comprehensively consider multiple factors in complex scenarios. Secondly, when processing video surveillance data, existing systems mostly rely on simple image processing technologies, such as motion detection, and cannot accurately identify whether workers are wearing personal protective equipment such as helmets and reflective vests, nor can they accurately determine whether high-altitude workers are wearing safety belts as required. In addition, current safety management systems mostly focus on post-event analysis, that is, finding out the cause through retrospective analysis after the accident occurs, rather than effective prevention before the accident occurs. This passive management method is obviously not conducive to fundamentally reducing the accident rate. Therefore, how to shift from the existing post-event management model to pre-event prevention and in-event control has become an urgent problem to be solved. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a construction safety risk assessment method and system to solve the problems of incomplete data processing, inaccurate risk assessment and lack of targeted preventive measures in existing construction safety risk assessment methods.

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

[0007] In a first aspect, an embodiment of the present invention provides a construction safety risk assessment method, which comprises:

[0008] Collect multi-dimensional data from the construction site, pre-process the multi-dimensional data and mark the workers' behaviors to form a safety hazard database;

[0009] Identify unsafe behavior patterns of workers by building a neural network model;

[0010] Receive data streams from the construction site in real time, dynamically adjust risk assessment parameters according to current construction status and environmental changes, predict potential safety risks, and obtain prediction results;

[0011] generating corresponding preventive measures suggestions based on the prediction results;

[0012] Key indicators and abnormal conditions are displayed through the monitoring interface, and detailed analysis reports are generated.

[0013] As a preferred solution of the construction safety risk assessment method of the present invention, wherein: the multi-dimensional data includes worker location data, worker action data, environmental data, equipment status data and video monitoring data;

[0014] The environmental data includes temperature, humidity, dust concentration and noise level.

[0015] As a preferred solution of the construction safety risk assessment method of the present invention, the multi-dimensional data is preprocessed and the worker behaviors are marked to form a safety hazard database, which includes the following steps:

[0016] Perform outlier detection on the speed of worker position change based on worker position data, and use time interpolation to repair lost outliers;

[0017] Use Kalman filter to smooth the skeleton point detection results in video surveillance data;

[0018] Use the interquartile range method to remove outliers from environmental data and equipment status data;

[0019] Use the nearest neighbor time alignment method to align and synchronize the timestamps of all data;

[0020] Use convolutional neural networks combined with posture recognition technology to extract workers’ skeletal features from video surveillance data, analyze workers’ movement types, and generate labels;

[0021] Based on the comprehensive analysis of video surveillance data and equipment status data, standardized behavior labels for workers are formed;

[0022] The labeled worker behavior and environmental data are stored in the database to form a comprehensive safety hazard database.

[0023] As a preferred solution of the construction safety risk assessment method of the present invention, the following steps are included in constructing a neural network model to identify unsafe behavior patterns of workers:

[0024] Use the data in the labeled safety hazard database as training data for the model;

[0025] Use convolutional neural networks to extract spatial features of images from the annotated worker action types;

[0026] Use long short-term memory networks to extract temporal features of worker behaviors from labeled standardized behaviors of workers;

[0027] The convolutional neural network and the long short-term memory network are fused through a fully connected layer to train the neural network model to identify unsafe behavior patterns of workers.

[0028] As a preferred solution of the construction safety risk assessment method of the present invention, the data stream from the construction site is received in real time, and the risk assessment parameters are dynamically adjusted according to the current construction status and environmental changes. Predicting potential safety risks includes the following steps:

[0029] Through the sensor network and video surveillance system at the construction site, real-time collection of worker location data, worker movement data, environmental data, equipment status data and video surveillance data;

[0030] According to the historical accident cases at the construction site, the frequency of different types of accidents is counted, high-risk accident types are identified, the time distribution of accidents in a day is analyzed, high-risk time periods are identified, the location distribution of accidents at the construction site is analyzed, and high-risk areas are identified;

[0031] Analyze the correlation between current environmental data, equipment status data, worker location data, and motion data and the occurrence of accidents based on current conditions;

[0032] According to the analysis results, a composite function is set to predict potential security risks. The expression is:

[0033]

[0034] Where P(t) is the risk score at time t, R(t) is the worker location data at time t, A(t) is the worker action data at time t, E(t) is the environment data at time t, M(t) is the equipment status data at time t, V(t) is the video surveillance data at time t, λ is the adjustment factor of the video surveillance data affecting the risk score, and V0 is the baseline value of the video surveillance data;

[0035] Set the state space to worker location data, worker motion data, environmental data, equipment status data, video surveillance data, past accident records, worker historical behavior data, current task type, and work stage;

[0036] The action space is set to adjust the weights of each factor and the adjustment factor of the video surveillance data affecting the risk score;

[0037] The reward function is set to give positive rewards when a potential security risk is successfully predicted and avoided, negative rewards when a security incident occurs due to a failure to predict in time, and neutral rewards when the system operates normally without a security incident.

[0038] Select deep reinforcement learning algorithms suitable for discrete action space and continuous action space, evaluate the quality of each action and optimize the strategy through policy gradient method;

[0039] Continuously collect data on the current state, the action chosen, the reward obtained, and the new state after performing the action;

[0040] Keep trying different actions and adjust strategies based on reward feedback;

[0041] Based on the current state, the policy network selects the optimal action and adjusts the risk assessment parameters;

[0042] Recalculate the risk score using the adjusted risk assessment parameters.

[0043] As a preferred solution of the construction safety risk assessment method of the present invention, generating corresponding preventive measures suggestions according to the risk assessment results includes the following steps:

[0044] According to the calculated risk score results, the risks are divided into three levels: low risk, medium risk and high risk;

[0045] When the risk is low, no special intervention is required. Through mobile applications, workers are reminded to pay attention to safety, strictly follow standard operating procedures and conduct risk assessments in real time. Targeted operation guidance and safety training resources are provided according to the age and experience of workers;

[0046] When the risk is medium, the system reminds workers to wear protective equipment through voice and provides personalized protection suggestions based on the workers' health conditions;

[0047] When the risk is high, the system automatically sounds an alarm, notifying the site manager to suspend related operations and requiring safety managers to inspect the site until it is restored to a safe state, providing emergency response measures based on the workers' specific circumstances.

[0048] As a preferred solution of the construction safety risk assessment method of the present invention, wherein: displaying key indicators and abnormal conditions through a monitoring interface and generating a detailed analysis report includes the following steps:

[0049] Build a monitoring interface based on HTML5 Web interface combined with WebSocket real-time data push technology;

[0050] The monitoring interface displays the worker’s risk score, real-time speed, changes in bone position points, environmental data, and equipment vibration frequency;

[0051] When the collected data exceeds the threshold, the system issues an abnormal warning in real time and automatically generates a risk score time series graph;

[0052] Generate detailed analysis reports based on workers’ historical risk scores, abnormal event records, and preventive measures recommendations.

[0053] In a second aspect, the present invention provides a construction safety risk assessment system, comprising:

[0054] The data collection and annotation module collects multi-dimensional data from the construction site, pre-processes the multi-dimensional data and annotates the workers' behaviors to form a safety hazard database;

[0055] The safety identification module identifies unsafe behavior patterns of workers by building a neural network model;

[0056] The risk assessment module receives data streams from the construction site in real time, dynamically adjusts risk assessment parameters according to the current construction status and environmental changes, predicts potential safety risks, and obtains prediction results;

[0057] A risk prevention suggestion module generates corresponding preventive measures suggestions according to the prediction results;

[0058] The visualization module displays key indicators and abnormal conditions through the monitoring interface and generates detailed analysis reports.

[0059] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the construction safety risk assessment method as described in the first aspect of the present invention is implemented.

[0060] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the construction safety risk assessment method as described in the first aspect of the present invention is implemented.

[0061] The beneficial effects of the present invention are as follows: by collecting and preprocessing real-time data, data quality is ensured, laying the foundation for subsequent analysis, thereby improving the accuracy of safety hazard identification. A safety hazard database is established to record unsafe behaviors such as not wearing a safety helmet, which facilitates the learning and identification of intelligent models, enhances the system's intelligent decision-making capabilities, and improves workers' safety awareness. A neural network model is constructed to identify unsafe behavior patterns, and the accuracy and stability of the model are improved through parameter tuning, reducing false alarms and missed alarms. Dynamic assessment and timely response to safety risks at the construction site are achieved, and risk assessment parameters can be quickly adjusted according to environmental changes, potential risks can be predicted, and preventive measures can be recommended, effectively reducing the probability of safety accidents and improving construction efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0063] Figure 1 This is a flow chart of the construction safety risk assessment method in Example 1.

[0064] Figure 2 This is the risk score classification diagram in Example 1. DETAILED DESCRIPTION

[0065] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0066] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0067] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0068] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a construction safety risk assessment method, comprising the following steps:

[0069] S1. Multi-dimensional data includes worker location data, worker movement data, environmental data, equipment status data and video surveillance data; environmental data includes temperature, humidity, dust concentration and noise level.

[0070] Preprocessing multi-dimensional data and labeling worker behaviors to form a safety hazard database includes the following steps:

[0071] Outlier detection is performed on the speed of worker position changes based on worker position data, and the lost outliers are repaired using time interpolation. The skeleton point detection results in the video surveillance data are smoothed using a Kalman filter. The interquartile range method is used to remove outliers from the environment and equipment data.

[0072] Use the nearest neighbor time alignment method to align and synchronize the timestamps of all data. Use convolutional neural networks combined with posture recognition technology to extract the skeletal features of workers from video surveillance data, analyze the types of workers' movements, and generate labels. Based on the comprehensive analysis of video surveillance data and equipment status data, form standardized behavior labels for workers. Store the labeled worker behavior and environmental data in the database to form a comprehensive safety hazard database.

[0073] Furthermore, the specific operations of using convolutional neural network combined with posture recognition technology are as follows:

[0074] Deploy multiple high-definition cameras at the construction site to cover the main work areas and record videos of workers in different working conditions, including normal work and unsafe behaviors. Use video annotation tools to annotate the key skeletal points of each worker (such as head, shoulders, elbows, hands, waist, knees, ankles, etc.), and generate corresponding action labels for each annotated video clip, such as "normal walking", "carrying heavy objects", "climbing ladders", etc. Select a CNN architecture suitable for processing image data, input as a single-frame or multi-frame video image, and adjust the size according to actual conditions, such as 224x224 pixels. Use multiple convolutional layers to extract the spatial features of the image, reduce the dimension through the pooling layer to reduce the amount of calculation, and map the features extracted by the convolutional layer to the coordinates of the skeletal points. Use open source posture recognition libraries such as OpenPose (real-time multi-person key point detection system) to preliminarily extract the skeletal points of the workers in the video, and enhance the training data through methods such as rotation, scaling, and cropping. Use CNN to extract the coordinates of the skeletal points of the workers from the video frames. Use support vector machines (SVM) and random forests (Random Forest) or deep learning classifiers (such as LSTM) classify the extracted features and generate corresponding action labels based on the classification results, such as "normal walking", "carrying heavy objects", "climbing ladders", etc.; combined with video surveillance data, equipment status data and environmental data, a comprehensive analysis is performed to determine whether the worker's behavior complies with safety regulations. For example, if a worker works at a height but does not wear a safety belt, it is marked as "unsafe behavior".

[0075] S2. Constructing a neural network model to identify unsafe behavior patterns of workers includes the following steps: using the data in the labeled safety hazard database as training data for the model; using a convolutional neural network to extract the spatial features of the image from the labeled workers' action types; using a long short-term memory network to extract the temporal features of the workers' behavior from the labeled standardized behaviors of the workers; fusing the convolutional neural network with the long short-term memory network through a fully connected layer, training the neural network model, and identifying unsafe behavior patterns of workers.

[0076] Specifically, the operations for training the neural network model are: using the cross entropy loss function and the Adam optimizer, setting an appropriate batch size and a sufficient number of training rounds; using a portion of the data as a validation set, and regularly evaluating the performance of the model on the validation set to prevent overfitting; if the performance on the validation set does not improve within a certain number of rounds, stop training in advance; perform enhancement operations such as rotation, scaling, and cropping on the training data to increase the generalization ability of the model; using L2 regularization (weight decay) to prevent overfitting and setting regularization parameters; gradually reducing the learning rate during training, for example, multiplying the learning rate by 0.1 every 10 epochs; and regularly saving the best performing model during training.

[0077] S3, receiving data streams from the construction site in real time, dynamically adjusting risk assessment parameters according to current construction status and environmental changes, and predicting potential safety risks include the following steps:

[0078] Through the sensor network and video surveillance system at the construction site, real-time collection of worker location data, worker movement data, environmental data, equipment status data and video surveillance data;

[0079] According to the historical accident cases at the construction site, the frequency of different types of accidents is counted, high-risk accident types are identified, the time distribution of accidents in a day is analyzed, high-risk time periods are identified, the location distribution of accidents at the construction site is analyzed, and high-risk areas are identified;

[0080] Analyze the correlation between current environmental data, equipment status data, worker location data, and motion data and the occurrence of accidents based on current conditions. For example, high temperature and high humidity may increase the risk of heat stroke, high dust concentration may increase the risk of respiratory diseases, and the probability of accidents may be higher when the equipment failure rate is high. Frequently moving workers may be more prone to collision accidents, and irregular operations may cause equipment failures.

[0081] According to the analysis results, a composite function is set to predict potential security risks. The expression is:

[0082]

[0083] Where P(t) is the risk score at time t, R(t) is the worker location data at time t, A(t) is the worker action data at time t, E(t) is the environment data at time t, M(t) is the equipment status data at time t, V(t) is the video surveillance data at time t, λ is the adjustment factor of the video surveillance data affecting the risk score, and V0 is the baseline value of the video surveillance data;

[0084] Set the state space to worker location data, worker motion data, environmental data, equipment status data, video surveillance data, past accident records, worker historical behavior data, current task type, and work stage;

[0085] The action space is set to adjust the weights of each factor and the adjustment factor of the video surveillance data affecting the risk score;

[0086] The reward function is set to give positive rewards when a potential security risk is successfully predicted and avoided, negative rewards when a security incident occurs due to a failure to predict in time, and neutral rewards when the system operates normally without a security incident.

[0087] Select deep reinforcement learning algorithms suitable for discrete action space and continuous action space, evaluate the quality of each action and optimize the strategy through policy gradient method;

[0088] Continuously collect data on the current state, selected actions, obtained rewards, and the new state after performing the actions;

[0089] Keep trying different actions and adjust the strategy according to the reward feedback;

[0090] Based on the current state, the policy network selects the optimal action and adjusts the risk assessment parameters;

[0091] Recalculate the risk score using the adjusted risk assessment parameters.

[0092] Furthermore, the correlation analysis includes that high temperature and high humidity may increase the risk of heat stroke, high dust concentration may increase the risk of respiratory diseases, a high equipment failure rate indicates that the probability of an accident may be higher, workers who move frequently may be more prone to collision accidents, and non-standard operations may lead to equipment failures.

[0093] Specifically, a positive reward means that when a potential safety risk is successfully predicted and avoided, the reward is +1; a negative reward means that when a safety event occurs due to failure to predict in time, the reward is -1; a neutral reward means that when the operation is normal and no safety event occurs, the reward is 0.

[0094] S4. According to the risk assessment results, generating corresponding preventive measure suggestions includes the following steps.

[0095] Based on the calculated risk score results, divide the risks into three levels: low risk, medium risk, and high risk. When in a low-risk situation, no special intervention is required. Through the mobile application, remind workers to pay attention to safety, work strictly according to the standard operation procedures, and conduct real-time risk assessment. When in a medium-risk situation, the system reminds workers to wear protective equipment, such as safety helmets, seat belts, etc., through voice to reduce potential hazards. For example, if the sensor detects that a worker is not wearing safety equipment, the system will automatically send a reminder notice to notify the safety management personnel at the construction site to strengthen inspections and promptly handle potential hidden dangers. When in a high-risk situation, the system automatically issues an alarm, notifies the site supervisor to suspend the relevant operation, and requires the safety management personnel to check the site until the safety state is restored.

[0096] Furthermore, the specific division of the risk score is as follows:

[0097] The value range of the risk score is 0 ≤ P(t) < +∞.

[0098] When P(t) = 0, it means that the worker is in a completely safe state, the environment and equipment are normal, and the worker's behavior has no risks.

[0099] When 0 < P(t) < 0.5, it means that the worker is in a low-risk state, no additional safety measures are required, but should remain vigilant.

[0100] When 0.5≤P(t)<1.0, it means that the worker is at medium risk and it is recommended that the worker wear protective equipment and strengthen on-site safety monitoring;

[0101] When P(t) ≥ 1.0, it indicates that the workers are in a high-risk state and an alarm needs to be sounded immediately, construction needs to be suspended, and emergency preventive measures need to be taken.

[0102] S5. Display key indicators and abnormal conditions through the monitoring interface and generate a detailed analysis report including the following steps:

[0103] The monitoring interface is built based on the HTML5 Web interface combined with WebSocket real-time data push technology; the worker's risk score, real-time speed, changes in bone position points, environmental data and equipment vibration frequency are displayed in the monitoring interface; when the collected data exceeds the threshold, the system issues an abnormal warning in real time and automatically generates a risk score time series graph; the worker's historical risk score, abnormal event records and preventive measures are recorded to generate a detailed analysis report.

[0104] This embodiment also provides a construction safety risk assessment system, including: a data acquisition module, which collects multi-dimensional data from the construction site; a preprocessing module, which preprocesses the collected data and marks the workers' behavior to form a safety hazard database; a safety identification module, which builds a neural network model to identify the unsafe behavior patterns of workers; a risk assessment module, which receives data streams from the construction site in real time, dynamically adjusts risk assessment parameters according to the current construction status and environmental changes, and predicts potential safety risks; a risk prevention suggestion module, which generates corresponding preventive measures suggestions based on the prediction results; a visualization module, which displays key indicators and abnormal conditions through a monitoring interface, and generates a detailed analysis report.

[0105] This embodiment also provides a computer device, which is suitable for the construction safety risk assessment method, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the construction safety risk assessment method proposed in the above embodiment.

[0106] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0107] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for implementing the construction safety risk assessment proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination of them, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0108] In summary, the present invention ensures data quality by: collecting and preprocessing real-time data, laying the foundation for subsequent analysis, and thus improving the accuracy of safety hazard identification. A safety hazard database is established to record unsafe behaviors such as not wearing a safety helmet, which is convenient for the learning and identification of intelligent models, enhances the intelligent decision-making ability of the system, and improves the safety awareness of workers. A neural network model is constructed to identify unsafe behavior patterns, and the accuracy and stability of the model are improved through parameter tuning, reducing false alarms and missed alarms. Dynamic assessment and timely response to safety risks at the construction site are achieved, and risk assessment parameters can be quickly adjusted according to environmental changes, potential risks can be predicted, and preventive measures can be recommended, effectively reducing the probability of safety accidents and improving construction efficiency.

[0109] Example 2, referring to Table 1, is the second example of the present invention. To further verify the technical solution of the present invention, experimental simulation data of a construction safety risk assessment method are provided.

[0110] A large construction site was selected as the test site. The site is currently constructing a high-rise residential building, which involves a large number of high-altitude operations, heavy machinery operations and other high-risk activities. This experiment was carried out for a month of continuous monitoring, during which a large amount of multi-dimensional data was collected, including but not limited to worker location data, worker movement data, environmental data, equipment status data and video surveillance data.

[0111] Before the experiment began, all sensors and cameras were calibrated to ensure the accuracy and reliability of data collection. For environmental data, multiple thermometers and hygrometers, dust concentration detectors, and noise level meters were installed in different areas of the construction site to more comprehensively reflect the on-site environmental conditions. At the same time, workers were equipped with safety helmets and smart bracelets with GPS positioning functions to track their location and health status in real time. In addition, multiple high-definition cameras were set up to capture workers' movements and changes in the surrounding environment.

[0112] Advanced data preprocessing technology was used during data collection. Specifically, outlier detection was performed on the speed of worker position changes, and a small number of lost outliers were repaired using time interpolation to ensure data continuity and integrity. Then, the Kalman filter was used to smooth the skeleton point detection results in the video surveillance data to improve the accuracy of motion recognition. For environmental and equipment data, reasonable upper and lower limits were set to effectively remove possible outliers. All collected data were timestamped and synchronized using the nearest neighbor time alignment method to ensure consistency of data from different sources.

[0113] Next, we used deep learning technology, especially convolutional neural networks combined with posture recognition technology, to extract the skeletal features of workers from the video data, analyze the types of workers' movements, and generate corresponding labels. Subsequently, we processed the workers' time series data through long short-term memory networks, combined with equipment status data and video data, and annotated the dynamic changes in the time series of workers' behaviors. These annotated data were integrated into a comprehensive safety hazard database, providing a solid foundation for subsequent risk assessment.

[0114] According to the calculated risk score results, the risks are divided into three levels: low, medium and high. For different levels of risk, corresponding preventive measures are recommended. When the risk is low, the mobile application reminds workers to pay attention to safety; when the risk is medium, the system will remind workers to wear protective equipment through voice and notify safety managers to strengthen inspections; when the risk is high, an alarm will automatically sound to notify the site manager to suspend related operations until safety is confirmed. At the same time, key indicators and abnormal conditions are displayed through the monitoring interface, and detailed analysis reports are generated, providing strong support for safety management.

[0115] The details are shown in Table 1 below:

[0116] Table 1 Experimental data record table

[0117]

[0118] By analyzing the data in the above table, it can be clearly seen that the method of the present invention has significant advantages in improving the accuracy and effectiveness of construction safety risk assessment. In all six scenarios, the risk scores of the method of the present invention are lower than the risk scores of the prior art. For example, in scenario 1, the risk score of the prior art is 0.65, while the risk score of the present invention is 0.42; in scenario 4, the risk score of the prior art is 0.82, while the risk score of the present invention is 0.56. This shows that the method of the present invention can more accurately identify and assess potential safety risks in construction.

[0119] The ambient temperature, humidity, dust concentration and noise level data in the table reflect the environmental conditions in different construction scenarios. The method of the present invention can better adapt to environmental changes by dynamically adjusting the risk assessment parameters. For example, when the ambient temperature is high (such as scenario 2 and scenario 4), the risk score of the method of the present invention remains at a low level, indicating that it is more adaptable to high temperature environments.

[0120] By calculating the risk factors of workers’ work intensity and equipment status, the method of the present invention can more comprehensively evaluate workers’ work status and equipment operation status. For example, in scenario 4, although the ambient temperature is as high as 36°C, the humidity is 85%, and the dust concentration is 1.8 mg / m 3 , the noise level is 95dB, and the risk score of the method of the present invention is still low, indicating that it can effectively consider the influence of multiple factors.

[0121] Based on the risk assessment results, the method of the present invention can generate corresponding preventive measures in a timely manner. For example, in high-risk situations, the system will automatically issue an alarm to notify the site manager to suspend related operations and require the safety manager to inspect the site until it is restored to a safe state. Such timely intervention measures can significantly reduce the risk of accidents.

[0122] The content of the present invention not only improves the accuracy of risk assessment, but also enhances the response speed and adaptability of the system. By receiving data streams from the construction site in real time and dynamically adjusting risk assessment parameters, the method can quickly respond to environmental changes and promptly discover and deal with potential safety hazards, which is of great significance for protecting the lives of workers and reducing accident losses.

[0123] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A construction safety risk assessment method, characterized in that: include, Collect multi-dimensional data from the construction site, pre-process the multi-dimensional data and mark the workers' behaviors to form a safety hazard database; Identify unsafe behavior patterns of workers by building a neural network model; Receive data streams from the construction site in real time, dynamically adjust risk assessment parameters according to current construction status and environmental changes, predict potential safety risks, and obtain prediction results; generating corresponding preventive measures suggestions based on the prediction results; Key indicators and abnormal conditions are displayed through the monitoring interface, and detailed analysis reports are generated.

2. The construction safety risk assessment method according to claim 1, characterized in that: The multi-dimensional data includes worker location data, worker motion data, environmental data, equipment status data and video surveillance data; The environmental data includes temperature, humidity, dust concentration and noise level.

3. The construction safety risk assessment method according to claim 2, characterized in that: Preprocessing the multi-dimensional data and labeling the worker behaviors to form a safety hazard database includes the following steps: Perform outlier detection on the speed of worker position change based on worker position data, and use time interpolation to repair the lost outliers; Use Kalman filter to smooth the skeleton point detection results in video surveillance data; Use the interquartile range method to remove outliers from environmental data and equipment status data; Use the nearest neighbor time alignment method to align and synchronize the timestamps of all data; Use convolutional neural networks combined with posture recognition technology to extract workers’ skeletal features from video surveillance data, analyze workers’ movement types, and generate labels; Based on the comprehensive analysis of video surveillance data and equipment status data, standardized behavior labels for workers are formed; The labeled worker behavior and environmental data are stored in the database to form a comprehensive safety hazard database.

4. The construction safety risk assessment method according to claim 3, characterized in that: By building a neural network model, identifying workers' unsafe behavior patterns includes the following steps: Use the data in the labeled safety hazard database as training data for the model; Use convolutional neural networks to extract spatial features of images from the annotated worker action types; Use long short-term memory networks to extract temporal features of worker behaviors from labeled standardized behaviors of workers; The convolutional neural network and the long short-term memory network are fused through a fully connected layer to train the neural network model to identify unsafe behavior patterns of workers.

5. The construction safety risk assessment method according to claim 4, characterized in that: Receive data streams from the construction site in real time, dynamically adjust risk assessment parameters according to current construction status and environmental changes, and predict potential safety risks. The following steps are included: Through the sensor network and video surveillance system at the construction site, real-time collection of worker location data, worker movement data, environmental data, equipment status data and video surveillance data; According to the historical accident cases at the construction site, the frequency of different types of accidents is counted, high-risk accident types are identified, the time distribution of accidents in a day is analyzed, high-risk time periods are identified, the location distribution of accidents at the construction site is analyzed, and high-risk areas are identified; Analyze the correlation between current environmental data, equipment status data, worker location data, and motion data and the occurrence of accidents based on current conditions; According to the analysis results, a composite function is set to predict potential security risks. The expression is: Where P(t) is the risk score at time t, R(t) is the worker location data at time t, A(t) is the worker action data at time t, E(t) is the environment data at time t, M(t) is the equipment status data at time t, V(t) is the video surveillance data at time t, λ is the adjustment factor of the video surveillance data affecting the risk score, and V0 is the baseline value of the video surveillance data; Set the state space to worker location data, worker motion data, environmental data, equipment status data, video surveillance data, past accident records, worker historical behavior data, current task type, and work stage; The action space is set to adjust the weights of each factor and the adjustment factor of the video surveillance data affecting the risk score; The reward function is set to give positive rewards when a potential security risk is successfully predicted and avoided, negative rewards when a security incident occurs due to a failure to predict in time, and neutral rewards when the system operates normally without a security incident. Select deep reinforcement learning algorithms suitable for discrete action space and continuous action space, evaluate the quality of each action and optimize the strategy through policy gradient method; Continuously collect data on the current state, the action chosen, the reward obtained, and the new state after performing the action; Keep trying different actions and adjust strategies based on reward feedback; Based on the current state, the policy network selects the optimal action and adjusts the risk assessment parameters; Recalculate the risk score using the adjusted risk assessment parameters.

6. The construction safety risk assessment method according to claim 5, characterized in that: According to the prediction results, generating corresponding preventive measures includes the following steps: According to the calculated risk score results, the risks are divided into three levels: low risk, medium risk and high risk; When the risk is low, no special intervention is required. Through mobile applications, workers are reminded to pay attention to safety, strictly follow standard operating procedures and conduct risk assessments in real time. Targeted operation guidance and safety training resources are provided according to the age and experience of workers; When the risk is medium, the system reminds workers to wear protective equipment through voice and provides personalized protection suggestions based on the workers' health conditions; When the risk is high, the system automatically sounds an alarm, notifying the site manager to suspend related operations and requiring safety managers to inspect the site until it is restored to a safe state, providing emergency response measures based on the workers' specific circumstances.

7. The construction safety risk assessment method according to claim 6, characterized in that: Displaying key indicators and abnormal conditions through the monitoring interface and generating detailed analysis reports include the following steps: Build a monitoring interface based on HTML5 Web interface combined with WebSocket real-time data push technology; The monitoring interface displays the worker’s risk score, real-time speed, changes in bone position points, environmental data, and equipment vibration frequency; When the collected data exceeds the threshold, the system issues an abnormal warning in real time and automatically generates a risk score time series graph; Generate detailed analysis reports based on workers’ historical risk scores, abnormal event records, and preventive measures recommendations.

8. A construction safety risk assessment system, based on the construction safety risk assessment method according to any one of claims 1 to 7, characterized in that: include, The data collection and annotation module collects multi-dimensional data from the construction site, pre-processes the multi-dimensional data and annotates the workers' behaviors to form a safety hazard database; The safety identification module identifies unsafe behavior patterns of workers by building a neural network model; The risk assessment module receives data streams from the construction site in real time, dynamically adjusts risk assessment parameters according to the current construction status and environmental changes, predicts potential safety risks, and obtains prediction results; A risk prevention suggestion module generates corresponding preventive measures suggestions according to the prediction results; The visualization module displays key indicators and abnormal conditions through the monitoring interface and generates detailed analysis reports.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the construction safety risk assessment method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the construction safety risk assessment method according to any one of claims 1 to 7 are implemented.

Images