Construction potential safety hazard automatic identification method and system based on deep learning

Abstract

The invention relates to a construction safety management technology, and discloses a construction potential safety hazard automatic identification method and system based on deep learning to improve the efficiency and accuracy of construction potential safety hazard identification. The method comprises the steps of performing feature extraction on historical construction anomaly information to obtain time period features, personnel behavior features, environment parameter features and equipment operation features, and training a preset construction anomaly hidden danger model to obtain a construction anomaly hidden danger model. And processing the personnel behavior information, the equipment operation information and the environment parameter information to obtain construction precursor data. Inputting the construction precursor data into the construction abnormal hidden danger model for identification to obtain a hidden danger identification result, if at least one abnormity exists in the hidden danger identification result, determining an area position with the hidden danger abnormity, performing hidden danger grade analysis on the area position to obtain a hidden danger abnormity grade, and if the hidden danger abnormity grade meets a preset hidden danger grade, determining that the hidden danger abnormity exists. And if yes, generating hidden danger abnormal early warning.

Description

Technical Field

[0001] The present invention relates to construction safety management technology, and in particular to a method and system for automatically identifying construction safety hazards based on deep learning. Background Art

[0002] As a key pillar of the national economy, the construction industry's development is directly linked to infrastructure development and socioeconomic progress. However, the industry has long faced severe challenges stemming from high risks and high accident rates, making construction site safety management a top priority.

[0003] Construction sites are a remarkably complex and dynamic environment, involving large numbers of people working across multiple sectors, a wide variety of machinery and equipment, densely stacked construction materials, and constantly changing work processes. These intertwined and interdependent factors contribute to the diversity, concealment, and sudden nature of construction safety hazards. Common safety hazards include falls from heights, impacts, mechanical injuries, electric shocks, and collapses. Once these hazards trigger accidents, they often result in serious casualties and property damage.

[0004] Currently, most construction companies still rely on the traditional manual inspection model to identify safety hazards. This model relies primarily on safety managers to conduct regular or irregular inspections of construction sites based on their personal experience and expertise. However, this approach has insurmountable limitations: (1) Low recognition efficiency: Construction sites are usually vast and have scattered work points, making it difficult for safety management personnel to achieve comprehensive coverage in a short period of time. Inspection blind spots and missed inspections are very likely to occur, especially for large-scale buildings or complex work scenarios. The timeliness of manual inspections cannot meet the needs of real-time safety monitoring.

[0005] (2) Identification accuracy is subject to subjective factors: Different managers have different experience reserves and professional levels, and their judgments on the same hidden danger may be inconsistent, making it difficult to ensure the accuracy and reliability of hidden danger identification.

[0006] (3) High labor costs and work intensity: To cover all operating areas, companies need to invest a large amount of manpower, resulting in high management costs. Furthermore, inspectors are often exposed to complex outdoor environments for long periods of time, facing high workloads and prone to fatigue, which further impacts inspection effectiveness.

[0007] These limitations directly lead to low efficiency and accuracy in identifying construction safety hazards, making it difficult to adapt to the refined and intelligent safety management requirements of modern construction. There is an urgent need for an efficient and accurate automatic identification technology to make up for the shortcomings of traditional models. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to provide a method and system for automatically identifying construction safety hazards based on deep learning, so as to improve the efficiency and accuracy of identifying construction safety hazards.

[0009] The technical solution adopted by the present invention to solve the above technical problems is: In one aspect, the present invention provides a method for automatically identifying construction safety hazards based on deep learning, comprising: Obtaining personnel behavior information, equipment operation information, and environmental parameter information. The equipment operation information includes the operation information of each device itself and the interactive operation information of related devices. The environmental parameter information refers to the environmental information of the area where the device is located. Collecting historical construction abnormality information and the cause of each construction abnormality event in the historical construction abnormality information, and performing feature extraction on the historical construction abnormality information to obtain time period features and personnel behavior features, environmental parameter features, and equipment operation features corresponding to the time period features, wherein the time period features include a first time period before each construction abnormality event occurs, a time period during the occurrence, and a second time period after the occurrence; Training a preset construction abnormality hidden danger model based on the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain a construction abnormality hidden danger model; Processing the personnel behavior information, the equipment operation information, and the environmental parameter information according to the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain construction precursor data; Inputting the construction precursor data into the construction abnormality hidden danger model for identification to obtain hidden danger identification results; Determining whether there is at least one hidden danger anomaly in the hidden danger identification result; if so, determining the location of the region where the hidden danger anomaly exists, and performing a hidden danger weight analysis on the region location to obtain a hidden danger coefficient corresponding to the region location; Determine the hidden danger level based on the hidden danger coefficient corresponding to the regional location and the construction abnormality data and the cause of the construction abnormality in the hidden danger identification result to obtain the hidden danger abnormality level; Determine whether the hidden danger abnormality level meets the preset abnormality level. If so, generate a hidden danger abnormality warning based on the construction abnormality data and the cause of the construction abnormality.

[0010] Furthermore, the training of a preset construction abnormality hidden danger model based on the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain the construction abnormality hidden danger model includes: determining a first behavior distinguishing feature, a first environmental parameter feature difference, and a first device operation feature difference based on the personnel behavior features, environmental parameter features, and device operation features corresponding to the second time period and the personnel behavior features, environmental parameter features, and device operation features corresponding to the first time period; Determine a second behavior distinguishing feature, a second environmental parameter feature difference, and a second device operation feature difference based on the personnel behavior features, environmental parameter features, and device operation features corresponding to the second time period and the personnel behavior features, environmental parameter features, and device operation features corresponding to the time period of the occurrence process; Sorting out construction abnormality trends based on the first behavioral distinguishing features, the first environmental parameter characteristic differences, and the first equipment operation characteristic differences to obtain precursor dynamic data corresponding to each construction abnormality event; Sorting out construction anomalies based on the second behavior distinguishing feature, the second environmental parameter feature difference, and the second equipment operation feature difference to obtain construction anomaly data corresponding to each construction anomaly event; The precursor dynamic data, the construction abnormality data and the cause of the construction abnormality are respectively input into a preset construction abnormality hidden danger model according to time nodes for training to obtain a construction abnormality hidden danger model.

[0011] Furthermore, the personnel behavior information, the equipment operation information, and the environmental parameter information are processed according to the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain construction precursor data, including: Extracting features from the personnel behavior information, the equipment operation information, and the environmental parameter information to obtain real-time behavior features, real-time operation features, and real-time parameter features; Performing feature matching on the real-time parameter feature and the real-time operation feature with the environmental parameter feature and the equipment operation feature during the time period during which the time period feature occurs, respectively, to determine whether there is an operation feature match and / or a parameter feature match; if not, performing feature matching on the real-time behavior feature, the real-time operation feature, and the real-time parameter feature with the human behavior feature, the environmental parameter feature, and the equipment operation feature during a second time period after the time period feature occurs, respectively, to obtain a comparison behavior feature, a comparison parameter feature, and a comparison operation feature; Based on the comparison behavior feature, the comparison parameter feature and the comparison operation feature, feature comparison processing is performed on the real-time behavior feature, the real-time operation feature and the real-time parameter feature to obtain construction precursor data.

[0012] Furthermore, the performing feature comparison processing on the real-time behavior feature, the real-time operation feature, and the real-time parameter feature based on the comparison behavior feature, the comparison parameter feature, and the comparison operation feature to obtain construction precursor data includes: Performing behavioral feature distinction analysis on the compared behavioral features and the real-time behavioral features to obtain real-time behavioral distinction features; Perform parameter feature difference comparison on the real-time parameter feature and the comparison parameter feature to obtain a real-time environment parameter feature difference; Performing operation characteristic difference comparison between the real-time operation characteristic and the comparison operation characteristic to obtain a real-time operation parameter characteristic difference; The real-time behavior distinguishing features, the real-time environmental parameter characteristic differences, and the real-time operating parameter characteristic differences are trended to obtain construction precursor data.

[0013] Furthermore, performing a hidden danger weight analysis on the regional location to obtain a hidden danger coefficient corresponding to the regional location includes: Acquire an image of the construction area, and construct a three-dimensional construction model of the construction area based on the image of the construction area; Determining, based on the three-dimensional construction model, concentrated hazard areas that meet preset conditions and dispersed hazard areas that do not meet preset conditions, wherein the preset conditions are that a specific number of people are present and adjacent equipment is within a specific distance; The concentrated hidden danger area and the dispersed hidden danger area are respectively used as hidden danger area points, and the hidden danger area points are connected according to the time node sequence corresponding to the abnormal construction event to obtain a hidden danger trend curve; Perform regular extension analysis on the hidden danger trend curve to obtain a predicted trend curve of the probability of hidden danger anomalies occurring in future periods; Determine the future regional points corresponding to the predicted abnormal trend curve, and determine the hidden danger coefficients corresponding to different regional positions in the construction area based on the concentration and dispersion characteristics of the future regional points and the hidden danger regional points.

[0014] Furthermore, the regular extension analysis of the hidden danger trend curve is performed to obtain a prediction trend curve of the probability of hidden danger anomalies occurring in future periods, including: Determine, based on the hidden danger trend curve, the hidden danger development vector extending from different hidden danger area points to the next hidden danger area point and the abnormal feature set corresponding to each hidden danger area point; Determining a hidden danger vector feature based on the hidden danger development vector, and adding the hidden danger vector feature to the abnormal feature set to obtain a hidden danger feature set; Arranging each of the hidden danger feature sets in a matrix according to a time sequence to obtain a hidden danger feature matrix corresponding to each hidden danger feature set; Deducing the hidden danger feature matrix from the hidden danger feature set according to the time period to obtain multiple hidden danger point matrices in the future period; A future hidden danger feature set is determined based on multiple hidden danger point matrices, and the future hidden danger feature set is imported into the construction three-dimensional model to obtain a prediction trend curve of the probability of hidden danger anomalies occurring in future periods.

[0015] In a second aspect, the present invention further provides a system for automatically identifying construction safety hazards based on deep learning, comprising: An information acquisition module is used to obtain personnel behavior information, equipment operation information, and environmental parameter information. The equipment operation information includes the operation information of each device itself and the interactive operation information of related devices. The environmental parameter information is the environmental information of the area where the device is located; a collection and extraction module for collecting historical construction abnormality information and the cause of each construction abnormality event in the historical construction abnormality information, and performing feature extraction on the historical construction abnormality information to obtain time period features and personnel behavior features, environmental parameter features, and equipment operation features corresponding to the time period features, wherein the time period features include a first time period before each construction abnormality event occurs, a time period during the occurrence, and a second time period after the occurrence; A model training module is used to train a preset construction abnormality hidden danger model based on the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain a construction abnormality hidden danger model; an information processing module, configured to process the personnel behavior information, the equipment operation information, and the environmental parameter information according to the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain construction precursor data; a hidden danger identification module, configured to input the construction precursor data into the construction abnormal hidden danger model for identification, and obtain hidden danger identification results; a weight analysis module, configured to determine whether there is at least one hidden danger anomaly in the hidden danger identification result; if so, determine the location of the region where the hidden danger anomaly exists, and perform hidden danger weight analysis on the region to obtain a hidden danger coefficient corresponding to the region; a hidden danger determination module, configured to determine the hidden danger level based on the hidden danger coefficient corresponding to the regional location and the construction abnormality data and the cause of the construction abnormality in the hidden danger identification result, and obtain a hidden danger abnormality level; The hidden danger warning module is used to determine whether the hidden danger abnormality level meets the preset abnormality level. If it does, it generates a hidden danger abnormality warning based on the construction abnormality data and the cause of the construction abnormality.

[0016] Furthermore, the model training module, when training a preset construction abnormality hidden danger model based on the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain the construction abnormality hidden danger model, is specifically used to: determining a first behavior distinguishing feature, a first environmental parameter feature difference, and a first device operation feature difference based on the personnel behavior features, environmental parameter features, and device operation features corresponding to the second time period and the personnel behavior features, environmental parameter features, and device operation features corresponding to the first time period; Determine a second behavior distinguishing feature, a second environmental parameter feature difference, and a second device operation feature difference based on the personnel behavior features, environmental parameter features, and device operation features corresponding to the second time period and the personnel behavior features, environmental parameter features, and device operation features corresponding to the time period of the occurrence process; Sorting out construction abnormality trends based on the first behavioral distinguishing features, the first environmental parameter characteristic differences, and the first equipment operation characteristic differences to obtain precursor dynamic data corresponding to each construction abnormality event; Sorting out construction anomalies based on the second behavior distinguishing feature, the second environmental parameter feature difference, and the second equipment operation feature difference to obtain construction anomaly data corresponding to each construction anomaly event; The precursor dynamic data, the construction abnormality data and the cause of the construction abnormality are respectively input into a preset construction abnormality hidden danger model according to time nodes for training to obtain a construction abnormality hidden danger model.

[0017] Furthermore, the information processing module, when processing the personnel behavior information, the equipment operation information, and the environmental parameter information according to the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain construction precursor data, is specifically configured to: Extracting features from the personnel behavior information, the equipment operation information, and the environmental parameter information to obtain real-time behavior features, real-time operation features, and real-time parameter features; Performing feature matching on the real-time parameter feature and the real-time operation feature with the environmental parameter feature and the equipment operation feature during the time period during which the time period feature occurs, respectively, to determine whether there is an operation feature match and / or a parameter feature match; if not, performing feature matching on the real-time behavior feature, the real-time operation feature, and the real-time parameter feature with the human behavior feature, the environmental parameter feature, and the equipment operation feature during a second time period after the time period feature occurs, respectively, to obtain a comparison behavior feature, a comparison parameter feature, and a comparison operation feature; Based on the comparison behavior feature, the comparison parameter feature and the comparison operation feature, feature comparison processing is performed on the real-time behavior feature, the real-time operation feature and the real-time parameter feature to obtain construction precursor data.

[0018] Furthermore, the weight analysis module, when performing a hidden danger weight analysis on the regional location to obtain a hidden danger coefficient corresponding to the regional location, is specifically used to: Acquire an image of the construction area, and construct a three-dimensional construction model of the construction area based on the image of the construction area; Determining, based on the three-dimensional construction model, concentrated hazard areas that meet preset conditions and dispersed hazard areas that do not meet preset conditions, wherein the preset conditions are that a specific number of people are present and adjacent equipment is within a specific distance; The concentrated hidden danger area and the dispersed hidden danger area are respectively used as hidden danger area points, and the hidden danger area points are connected according to the time node sequence corresponding to the abnormal construction event to obtain a hidden danger trend curve; Perform regular extension analysis on the hidden danger trend curve to obtain a predicted trend curve of the probability of hidden danger anomalies occurring in future periods; Determine the future regional points corresponding to the predicted abnormal trend curve, and determine the hidden danger coefficients corresponding to different regional positions in the construction area based on the concentration and dispersion characteristics of the future regional points and the hidden danger regional points.

[0019] In a third aspect, the present invention also provides an electronic device comprising a memory and a processor; a computer program is stored in the memory; when the processor executes the computer program, the above-mentioned method for automatic identification of construction safety hazards based on deep learning is implemented.

[0020] In a fourth aspect, the present invention also provides a computer-readable storage medium on which a computer program is stored; when the computer program is executed by a processor, the above-mentioned method for automatic identification of construction safety hazards based on deep learning is implemented.

[0021] The beneficial effects of the present invention are: (1) Improve the comprehensiveness and accuracy of hidden danger identification: By simultaneously acquiring personnel behavior information, equipment operation information (including interactive operation information of the equipment itself and related equipment) and environmental parameter information, the key elements of the construction site are comprehensively covered, and an identification system is constructed from multiple dimensions such as personnel operation, equipment status, and environmental impact. This avoids missed detection problems caused by a single information source and greatly improves the accuracy of hidden danger identification.

[0022] (2) Realize intelligent and automated identification of hidden dangers: A construction anomaly and hidden danger model is constructed based on deep learning technology. The model is trained through historical construction anomaly data, enabling it to have the ability to autonomously learn and identify hidden dangers, replacing the traditional manual inspection mode, reducing interference from human subjective factors, realizing the automatic identification of construction safety hazards, and significantly improving identification efficiency.

[0023] (3) Strengthen the forward-looking early warning capability of hidden dangers: By extracting the time period characteristics of historical abnormal events (before, during, and after occurrence) and analyzing the dynamic data of abnormal event precursors, the model can generate construction precursor data based on real-time data, identify potential hidden dangers in advance, and issue early warnings, thereby buying time for hidden danger treatment and reducing the probability of accidents.

[0024] (4) Accurately locate hidden danger areas and scientifically assess risk levels: The hidden danger areas are located by combining the three-dimensional construction model, and the hidden danger coefficient is calculated through hidden danger weight analysis. The hidden danger level is determined by comprehensively combining the hidden danger coefficient, abnormal data and causes, making the risk assessment more objective and scientific, facilitating the formulation of targeted disposal strategies and improving the level of refined safety management.

[0025] (5) Reduce safety management costs and workload: The automated identification and early warning mechanism reduces reliance on manual inspections, lowers manpower input costs, and avoids fatigue and omissions caused by high-intensity work by inspectors, thereby improving overall safety management efficiency and adapting to the intelligent safety management needs of modern construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a flow chart of a method for automatically identifying construction safety hazards based on deep learning in one embodiment of the present invention.

[0027] Figure 2 It is a structural diagram of a system for automatically identifying construction safety hazards based on deep learning in one embodiment of the present invention.

[0028] Figure 3 It is a principle block diagram of an electronic device in one embodiment of the present invention. DETAILED DESCRIPTION

[0029] The present invention aims to provide a method and system for automatically identifying construction safety hazards based on deep learning, improving the efficiency and accuracy of construction safety hazard identification. Its core concept is to achieve automatic identification and early warning of construction safety hazards by building and training a deep learning model. Specifically, historical construction anomaly information and its corresponding causes are collected. Features covering the three time periods before, during, and after the anomaly are extracted, including human behavior characteristics, environmental parameter characteristics, and equipment operation characteristics (information about the equipment itself and its interactions with associated equipment) corresponding to each time period. This is then used to train a pre-set construction anomaly hazard model, enabling it to identify hazards. In practical applications, the human behavior information, equipment operation information, and environmental parameter information to be identified are obtained. This real-time information is processed using the aforementioned features to generate construction precursor data, which is then input into the trained model for identification. If a hazard anomaly is identified, the location of the abnormal area is determined. A three-dimensional construction model is constructed to analyze the distribution of hazards in the area (concentrated or dispersed). A hazard trend curve is generated and future trends are predicted. A hazard coefficient is calculated, and the hazard level is determined based on the anomaly data and cause. When the level meets preset criteria, a hazard warning message containing the anomaly data and cause is generated. In this way, the present invention forms a complete automatic identification system from data collection, model training to real-time identification, risk assessment and early warning.

[0030] In terms of specific implementation, the implementation ideas of the solution of the present invention include: first, by comprehensively acquiring personnel behavior information, equipment operation information (including the operation data of a single device itself and the interactive operation information between related devices) and environmental parameter information, it can completely cover the key elements of the construction site, thereby evaluating the construction status from multiple dimensions such as personnel operation, equipment status, and environmental impact, greatly improving the accuracy of discovering potential hidden dangers and enhancing construction safety.

[0031] At the same time, by collecting historical construction anomaly information and the corresponding anomaly causes, we can extract time period characteristics (i.e., the first time period before the anomaly occurs, the second time period during and after the anomaly occurs), as well as the corresponding personnel behavior characteristics, environmental parameter characteristics, and equipment operation characteristics of each time period. This time period division clearly illustrates the manifestations of abnormal events at different stages, facilitating analysis of their development process. The three types of characteristics, personnel, environment, and equipment, reveal the inherent causes of construction anomalies from different dimensions, providing key data support for the construction of a construction anomaly hazard model. This model can learn the characteristic patterns of various abnormal events, laying the foundation for subsequent accurate hazard identification.

[0032] On this basis, a pre-set model is trained using extracted human behavior, environmental parameters, and equipment operation characteristics to develop a model capable of identifying construction anomalies and hidden dangers. Because the input features encompass multiple factors contributing to abnormal events, the model can fully learn the correlations between different feature combinations and hidden dangers. Through continuous parameter optimization through extensive data training, the model ultimately achieves high recognition accuracy and generalization capabilities, enabling reliable identification of potential hazards in practical applications and ensuring construction safety.

[0033] In real-time application, real-time information on personnel behavior, equipment operation, and environmental parameters is first processed with reference to historical features. After extracting real-time features, they are matched and compared with historical features to generate construction precursor data. This process filters and analyzes real-time information based on patterns in historical data, extracting key information related to hidden dangers. This makes the data input into the model more targeted, effectively improving recognition efficiency and accuracy, and facilitating the timely detection of potential anomalies.

[0034] Inputting construction precursor data into the trained model yields hazard identification results. Because the precursor data contains key construction features, and the model understands the correlation between features and hazards, it can quickly determine if anomalies exist and output results. This provides clear direction for subsequent hazard management and helps prevent accidents by taking timely measures.

[0035] If the identified hidden dangers are abnormal, the system will first determine the location of the abnormal area and then perform a hidden danger weight analysis on that area to determine the hidden danger coefficient. Accurate regional positioning provides accurate location information for hidden danger treatment; the weight analysis comprehensively considers the impact of factors such as personnel, environment, and equipment in the area on the hidden dangers, allowing the hidden danger coefficient to objectively reflect the severity of the regional hidden dangers, providing a scientific basis for subsequent level determination, and facilitating more accurate risk assessment and targeted treatment.

[0036] Subsequently, a hazard anomaly level is determined by combining the hazard coefficient, the construction anomaly data from the identification results, and the corresponding causes. This level integrates the regional risk level, the specific manifestations of the hazards, and the causes, providing a clear standard for handling decisions, improving the efficiency and effectiveness of hazard handling and reducing construction risks.

[0037] When the level of hidden danger anomalies reaches a preset standard, the system generates a hidden danger anomaly warning based on the construction anomaly data and causes. This warning mechanism, based on level determination, is triggered only when the hidden danger reaches a certain level of severity. The warning information includes specific anomaly data and causes, making it more targeted and practical, prompting relevant personnel to pay attention to and address hidden dangers, preventing them from escalating.

[0038] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. It should also be noted that the term "and / or" appearing in the present application indicates that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " appearing in the present application, unless otherwise specified, generally indicates that the objects associated with each other are in an "or" relationship.

[0039] One embodiment of the present application provides a method for automatically identifying construction safety hazards based on deep learning, which is executed by an electronic device, which can be a server or a terminal device. The server can be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services; the terminal device can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc., but is not limited to this. The terminal device and the server can be directly or indirectly connected through wired or wireless communication. This embodiment of the application does not limit this. Figure 1 As shown, the method includes steps S10 to S17: Step S10: Obtain personnel behavior information, equipment operation information, and environmental parameter information.

[0040] The equipment operation information includes the operation information of each equipment itself and the interactive operation information of related equipment. The environmental parameter information is the environmental information of the area where the equipment is located.

[0041] For the purposes of this application, personnel behavior information refers to a collection of data related to personnel activities. This refers to information such as personnel's actions, operating habits, and movement trajectories within the construction area, captured through various sensors and monitoring devices. For example, in a construction workshop, cameras installed near equipment can record data such as the operator's operating process, operating frequency, and movement routes within the workshop. This information constitutes personnel behavior information. Equipment operation information refers to data reflecting the equipment's operating status and operational status. This information includes each device's own operational information and interactive operation information for related devices. The operational information for each device represents the various parameters and status of the device during independent operation, such as its startup time, operating duration, workload, and energy consumption. The interactive operation information for related devices represents the coordination between related devices during operation within a production process or system, such as the flow rate and speed of raw materials delivered from one device to another, as well as the sequence of operations and signal interaction between the two devices. Environmental parameter information represents data on the environmental conditions within the area where the device is located. It refers to the information obtained through various environmental monitoring devices such as temperature sensors, humidity sensors, light sensors, and air quality sensors, such as the temperature, humidity, light intensity, and air quality (such as dust concentration and harmful gas concentration) of the equipment's environment.

[0042] In the embodiments of the present application, to acquire human behavior information, multiple high-definition cameras are first installed in the human activity area to ensure coverage of all areas where people are active. Image recognition technology is then used to analyze the video images captured by the cameras to identify information such as human behavior, movement, and location. The recognition results are then stored in a database for subsequent query and analysis. To acquire device operation information, a data acquisition module is first installed on each device to collect the device's own operating parameters, such as operating time and workload. For related devices, a communication network is then established between the devices, using a specific communication protocol (such as Modbus or Profinet) to enable data exchange between the devices and acquire interactive operation information. The collected device operation information is then transmitted to a data processing center via wired or wireless means for storage and processing. To acquire environmental parameter information, appropriate environmental monitoring devices, such as temperature sensors and humidity sensors, are selected based on the environmental characteristics of the area where the devices are located. These sensors are then installed in appropriate locations. The sampling frequency and data transmission method of the monitoring devices are then configured to ensure real-time and accurate acquisition of environmental parameter information. The data collected by the monitoring devices is then transmitted to the data processing center for data cleaning, analysis, and storage.

[0043] Step S11. Collect historical construction abnormality information and the construction abnormality cause corresponding to each construction abnormality event in the historical construction abnormality information, and perform feature extraction on the historical construction abnormality information to obtain time period features and personnel behavior features, environmental parameter features, and equipment operation features corresponding to the time period features.

[0044] The time period characteristics include a first time period before each abnormal construction event occurs, a time period during the occurrence, and a second time period after the occurrence.

[0045] For the embodiments of the present application, historical construction anomaly information refers to a data set formed by situations that occurred in past construction projects that were inconsistent with normal construction processes, standards or expectations. It refers to a detailed record of various abnormal conditions in the past construction process. Abnormal conditions include delayed construction progress, substandard construction quality, safety accidents, etc. Construction abnormal events refer to specific events that deviate from the normal construction state during the construction process. They are used to indicate unexpected situations caused by various factors during the construction process. These situations will have an adverse impact on the smooth progress, quality, safety, etc. of the construction. Construction abnormality causes refer to various factors that trigger the occurrence of construction abnormal events. It is used to indicate the root cause of abnormal conditions in construction. The cause may be human factors, such as construction personnel's operational errors, improper management personnel's instructions, or environmental factors, such as bad weather, changes in geological conditions, or equipment factors, such as equipment failure, aging, etc. The time period feature is used to represent the characteristics of construction abnormal events at different time stages. It includes the first time period before each construction abnormal event occurs, the time period during the occurrence, and the second time period after the occurrence. The first time period before an abnormal event occurs can help analyze potential influencing factors before the event occurs; the time period during the event reflects the real-time situation at the time of the abnormal event; and the second time period after the event helps assess the impact of the abnormal event on subsequent construction and the parameters after normal construction. Personnel behavior characteristics represent the behavioral characteristics of construction personnel during a specific time period. They are information reflecting personnel behavior characteristics extracted by analyzing data on their operating actions, work attitudes, and collaboration during the construction process. Environmental parameter characteristics represent the characteristics of relevant parameters of the construction environment during a specific time period. They are information reflecting changes in environmental conditions extracted by monitoring parameters such as temperature, humidity, wind speed, light intensity, and air quality. For example, if the ambient temperature suddenly rises during an abnormal construction event, the trend and magnitude of this temperature change are environmental parameter characteristics. Equipment operation characteristics represent the characteristics of the operating status of construction equipment during a specific time period. They are information reflecting the equipment's operating status extracted by analyzing data such as equipment operating parameters (such as speed, power, load, etc.), operating time, and fault frequency.

[0046] The method for collecting specific historical construction abnormality information and the construction abnormality causes corresponding to each construction abnormality event in the historical construction abnormality information is the same as the method for obtaining personnel behavior information, equipment operation information and environmental parameter information in step S10. The difference is that according to the records of the construction abnormality events that have occurred and the time period, the two types of data information are bound with the personnel behavior information, equipment operation information and environmental parameter information to obtain the construction abnormality information and the construction abnormality causes.

[0047] Step S12: training a preset construction abnormality hidden danger model based on personnel behavior characteristics, environmental parameter characteristics, and equipment operation characteristics to obtain a construction abnormality hidden danger model.

[0048] Specifically, a first behavioral distinguishing feature, a first environmental parameter feature difference, and a first equipment operation feature difference are determined based on the personnel behavior characteristics, environmental parameter characteristics, and equipment operation characteristics corresponding to the second time period and the personnel behavior characteristics, environmental parameter characteristics, and equipment operation characteristics corresponding to the first time period. A second behavioral distinguishing feature, a second environmental parameter feature difference, and a second equipment operation feature difference are determined based on the personnel behavior characteristics, environmental parameter characteristics, and equipment operation characteristics corresponding to the second time period and the personnel behavior characteristics, environmental parameter characteristics, and equipment operation characteristics corresponding to the time period during the occurrence process. The first behavioral distinguishing feature, the first environmental parameter feature difference, and the first equipment operation feature difference are sorted for construction anomaly trends to obtain precursor dynamic data corresponding to each construction anomaly event. The second behavioral distinguishing feature, the second environmental parameter feature difference, and the second equipment operation feature difference are sorted for construction anomalies to obtain construction anomaly data corresponding to each construction anomaly event. The precursor dynamic data, construction anomaly data, and the cause of the construction anomaly are input into a preset construction anomaly hidden danger model according to time nodes for training to obtain a construction anomaly hidden danger model.

[0049] Specifically, the first behavioral distinguishing feature represents the difference in human behavior between the second time period and the first time period. This feature is extracted by comparing human behavior data before and after the abnormal event, reflecting behavioral changes. The first environmental parameter characteristic difference represents the difference in environmental parameters between the second time period and the first time period. This feature is calculated by calculating the difference between environmental parameters (such as temperature and humidity) between the two time periods, reflecting environmental changes. For example, if the ambient temperature drops by 2 degrees Celsius after the abnormal event compared to before the event, this difference is the first environmental parameter characteristic difference. The first equipment operation characteristic difference represents the difference in equipment operation between the second time period and the first time period. This feature is extracted by comparing equipment operation data (such as operating efficiency and fault frequency) between the two time periods. The second behavioral distinguishing feature represents the difference in human behavior between the second time period and the duration of the abnormal event. This feature is extracted by comparing human behavior data after and during the abnormal event. The second environmental parameter characteristic difference represents the difference in environmental parameters between the second time period and the duration of the abnormal event. This feature is calculated by calculating the difference between environmental parameters between the two time periods, reflecting environmental changes. The second equipment operation characteristic difference represents the difference in equipment operation characteristics between the second time period and the time period during which the event occurred. This information is extracted by comparing equipment operation data from the two time periods. Precursor dynamic data refers to information that precedes an abnormal event by analyzing the characteristic differences between the first and second time periods. It indicates trends in human behavior, environmental parameters, and equipment operation before a construction abnormality occurs, indicating an impending abnormality. For example, if equipment efficiency gradually decreases one week before a failure, this downward trend is considered precursor dynamic data. Construction abnormality data refers to information related to the abnormal event, obtained by analyzing the characteristic differences between the time period during which the event occurred and the second time period. It indicates the specific manifestations of the construction abnormality and the recovery after the event. The pre-configured construction abnormality hazard model is a pre-built deep learning model framework for identifying construction abnormality hazards. Based on professional knowledge and experience in the construction field, it is a preliminary model structure that receives precursor dynamic data, construction abnormality data, and the causes of construction abnormalities, and uses algorithms and rules to identify potential hazards.

[0050] In an embodiment of the present application, data analysis tools (such as Python's Pandas library) are used to align and preprocess the time period feature data for each construction anomaly event to ensure accurate correspondence between the data's time nodes. A feature difference calculation script is then compiled to calculate the feature differences between the second time period and the first time period, and between the second time period and the time period during the occurrence process, respectively. This yields the first and second behavioral distinguishing features, the environmental parameter feature differences, and the equipment operation feature differences. Statistical analysis methods (such as regression analysis and trend analysis) are then used to organize the feature difference data, extracting precursor dynamic data and construction anomaly data. This organized data and the causes of the construction anomaly are then input into a pre-set machine learning model (such as a decision tree model). The model parameters are then adjusted through training to produce a construction anomaly hazard model.

[0051] Step S13: Processing the personnel behavior information, equipment operation information and environmental parameter information according to the personnel behavior characteristics, environmental parameter characteristics and equipment operation characteristics to obtain construction precursor data.

[0052] Specifically, feature extraction is performed on personnel behavior information, equipment operation information, and environmental parameter information to obtain real-time behavior features, real-time operation features, and real-time parameter features. Feature matching is performed on the real-time parameter features and real-time operation features with the environmental parameter features and equipment operation features during the time period of the occurrence process to determine whether there is an operation feature match and / or parameter feature match. If not, feature matching is performed on the real-time behavior features, real-time operation features, and real-time parameter features with the personnel behavior features, environmental parameter features, and equipment operation features during the second time period after the occurrence of the time period to obtain comparison behavior features, comparison parameter features, and comparison operation features. Feature comparison processing is performed on the real-time behavior features, real-time operation features, and real-time parameter features based on the comparison behavior features, comparison parameter features, and comparison operation features to obtain construction precursor data.

[0053] Specifically, behavioral feature differentiation analysis is performed on the compared behavior features and the real-time behavior features to obtain real-time behavioral differentiation features. Real-time parameter features are then differentiated and compared with the compared parameter features to obtain real-time environmental parameter feature differences. Real-time operation features are then differentiated and compared with the compared operation features to obtain real-time operation parameter feature differences. Trend analysis is performed on the real-time behavioral differentiation features, the real-time environmental parameter feature differences, and the real-time operation parameter feature differences to obtain construction precursor data.

[0054] Step S14: Input the construction precursor data into the construction abnormality hidden danger model for identification to obtain the hidden danger identification result.

[0055] Specifically, the output of the construction anomaly hazard model, after processing and analyzing the input construction precursor data, is determined based on the hazard identification results. This result clearly indicates whether a hazard exists in the construction scenario corresponding to the input data, as well as the specific circumstances of the hazard, such as its type (equipment failure, human error, environmental safety risk, etc.), its location, expected timeframe for occurrence, and potential hazards. For example, a hazard identification result might indicate, "In the operating area of ​​tower crane No. 3, there is a potential for equipment brake system failure, which is expected to occur within the next 24 hours."

[0056] Step S15: Determine whether there is at least one hidden danger anomaly in the hidden danger identification result. If so, determine the regional location where the hidden danger anomaly exists, and perform hidden danger weight analysis on the regional location to obtain the hidden danger coefficient corresponding to the regional location.

[0057] Specifically, an image of the construction area is obtained, and a three-dimensional construction model of the construction area is constructed based on the image of the construction area. Based on the three-dimensional construction model, concentrated hidden danger areas that meet preset conditions and dispersed hidden danger areas that do not meet preset conditions are determined. The preset conditions are that a specific number of people are present and adjacent equipment is within a specific distance. The concentrated hidden danger areas and dispersed hidden danger areas are respectively used as hidden danger area points, and the hidden danger area points are connected according to the time node sequence corresponding to the abnormal construction events to obtain a hidden danger trend curve. The hidden danger trend curve is subjected to regular extension analysis to obtain a predicted trend curve of the probability of hidden danger anomalies occurring in the future period. The future regional points corresponding to the predicted abnormal trend curve are determined, and the hidden danger coefficients corresponding to different regional locations in the construction area are determined based on the concentration and dispersion characteristics of the future regional points and the hidden danger area points.

[0058] In the embodiments of this application, a construction area image refers to image data reflecting the actual scene of a construction area, captured by various image acquisition devices (such as cameras and drone-mounted cameras). These images can contain information such as the construction site layout, building structure, equipment placement, and personnel activities, and serve as the fundamental data source for constructing a 3D construction model. For example, panoramic images captured by drones during aerial photography of a large construction site constitute a construction area image. A 3D construction model is a virtual 3D spatial model constructed based on the construction area image using 3D modeling techniques (such as computer vision and 3D reconstruction algorithms). It accurately depicts the three-dimensional structure, spatial relationships, and object morphology of the construction area, providing intuitive visualization support for subsequent hazard analysis. For example, a 3D construction model clearly shows the layout of each floor within a building, the location of equipment, and the movement paths of personnel. Within a 3D construction model, areas that meet specific conditions (a certain number of personnel and a certain distance between adjacent equipment) are defined as concentrated hazard areas. These areas, due to the dense concentration of personnel and equipment, present high safety risks, such as collisions and chain reactions caused by equipment failures. Dispersed hazard areas, as opposed to concentrated hazard areas, are defined as areas that do not meet the aforementioned specific conditions. These areas have relatively dispersed personnel and equipment distribution, resulting in relatively low safety hazards, but may also present isolated safety issues, such as individual equipment failures or personnel misconduct. For example, the periphery of a construction site, where personnel and equipment are sparse but hazards such as improper material stacking exist, is considered a dispersed hazard area. Hazard area points mark concentrated and dispersed hazard areas in the 3D construction model. Each marked point represents a hazard area point, which contains information such as the location and type of hazard area (concentrated or dispersed), forming the basis for subsequent hazard analysis and prediction. For example, in the 3D construction model, different colored markers represent concentrated and dispersed hazard areas, respectively. A hazard trend curve is formed by connecting hazard area points according to the time sequence of corresponding construction abnormal events. This curve reflects the changes in hazard areas over time, including changes in location and number, and helps analyze hazard trends. For example, a hazard trend curve can be used to observe the gradual expansion or shift of a concentrated hazard area over time. The predicted trend curve is an extended analysis of the hidden danger trend curve. Based on historical data and trend patterns, it predicts the changing probability of hidden danger anomalies in future cycles. The resulting curve is called the predicted trend curve. This curve provides forward-looking guidance for construction safety management, helping to take preventive measures. Future regional points are determined based on the predicted trend curve to identify regional points with hidden dangers in the future cycle.The hidden danger coefficient is determined for different locations within the construction area based on the future location of the area and the concentration and dispersion of the hidden danger points. The hidden danger coefficient is a quantitative indicator used to measure the likelihood of hidden dangers in different areas. A higher coefficient indicates a greater hidden danger risk. For example, the hidden danger coefficient is higher in areas with concentrated hidden dangers, while it is relatively lower in areas with dispersed hidden dangers.

[0059] For the embodiment of the present application, the specific calculation method of the hidden danger coefficient includes: (1) Determine the quantitative indicators of the degree of centralization and decentralization: Concentration Index (C): For each potential hazard area, count the number of personnel and equipment within a certain radius (e.g., within a 50-meter radius). The greater the number of personnel, the greater the number of equipment, and the closer the distance between them, the higher the concentration. For example, set the weight of personnel to 0.6 and the weight of equipment to 0.4, and use linear weighting to calculate the concentration index. If an area has 20 personnel and 10 pieces of equipment, with the equipment distances closely spaced, the personnel score is (20 / maximum capacity of the area) × 0.6 + the equipment score is (10 / maximum capacity of the area) × 0.4 = the concentration index. (Assuming the maximum capacity is 50 personnel and the maximum capacity is 20 equipment, the personnel score is 0.4, the equipment score is 0.5, and the concentration index is 0.4 × 0.6 + 0.5 × 0.4 = 0.44.)

[0060] Decentralization (D): In contrast to centralization, a high degree of decentralization indicates that personnel and equipment are more sparsely distributed. The decentralization index can be calculated as 1 minus the centralization index: D = 1 - C.

[0061] (2) Setting the basic hidden danger coefficient: The basic hidden danger coefficient for concentrated hidden danger areas is set to 1.5, and the basic hidden danger coefficient for dispersed hidden danger areas is set to 0.8.

[0062] (3) Calculate the final hidden danger coefficient: For each region, the basic hidden danger coefficient is adjusted based on its concentration and dispersion index. For example, using linear interpolation, the final hidden danger coefficient (H) = basic hidden danger coefficient × (1 + adjustment coefficient × (concentration index - baseline value)). The baseline value can be set based on actual conditions, such as 0.5. The adjustment coefficient can be set based on sensitivity to concentration and dispersion, such as 0.3. If the concentration index of a concentrated hidden danger area is 0.7, the basic hidden danger coefficient is 1.5, the baseline value is 0.5, and the adjustment coefficient is 0.3, then the final hidden danger coefficient H = 1.5 × (1 + 0.3 × (0.7 - 0.5)) = 1.5 × (1 + 0.06) = 1.59.

[0063] Specifically, the hazard development vector extending from different hazard area points to the next hazard area point and the abnormal feature set corresponding to each hazard area point are determined based on the hazard trend curve. The hazard vector features are determined based on the hazard development vectors, and the hazard vector features are added to the abnormal feature set to obtain a hazard feature set. Each hazard feature set is arranged in a matrix according to time sequence to obtain a hazard feature matrix corresponding to each hazard feature set. The hazard feature matrix is ​​deduced from the hazard feature set according to time periods to obtain multiple hazard point matrices in the future period. Based on the multiple hazard point matrices, the future hazard feature set is determined, and the future hazard feature set is imported into the three-dimensional construction model to obtain a predicted trend curve for the probability of hazard anomalies occurring in the future period.

[0064] Step S16: Determine the hidden danger level based on the hidden danger coefficient corresponding to the regional location and the construction abnormality data and the cause of the construction abnormality in the hidden danger identification result to obtain the hidden danger abnormality level.

[0065] In the embodiment of the present application, the hidden danger coefficient is divided into different intervals, corresponding to different basic hidden danger level tendencies.

[0066] For example: hidden danger coefficient H∈[0,0.3): low hidden danger tendency, corresponding to the basic hidden danger level of level one (minor hidden danger); hidden danger coefficient H∈[0.3,0.7): medium hidden danger tendency, corresponding to the basic hidden danger level of level two (general hidden danger); hidden danger coefficient H∈[0.7,1): high hidden danger tendency, corresponding to the basic hidden danger level of level three (serious hidden danger); then match the hidden danger coefficients obtained in step S15 with the corresponding interval coefficients respectively to obtain the basic hidden danger level.

[0067] The construction anomaly data and the cause of the construction anomaly are then input into the preset construction anomaly level determination criteria for comparison, resulting in the corresponding construction hazard level. The preset construction anomaly level determination criteria is a pre-established system of rules for classifying construction anomalies and determining hazard levels. It clarifies the hazard levels corresponding to different combinations of construction anomaly data and causes, and serves as the basis for grading construction anomalies. The construction hazard level is determined based on the construction anomaly data and causes, according to preset standards, to indicate the severity of the construction hazard. It is typically divided into Level 1 (minor hazard), Level 2 (general hazard), Level 3 (serious hazard), and Level 4 (particularly serious hazard), and is used to indicate the impact of the construction hazard on construction safety, progress, and quality.

[0068] The foundation hidden danger level and the construction hidden danger level are cumulatively calculated to obtain the hidden danger abnormality level.

[0069] Step S17: Determine whether the hidden danger abnormality level meets the preset abnormality level. If so, generate a hidden danger abnormality warning based on the construction abnormality data and the cause of the construction abnormality.

[0070] Specifically, the preset anomaly level is a pre-set reference level used to determine whether a potential hazard anomaly level warrants a warning. This level is determined based on factors such as the project's characteristics, safety requirements, and historical experience, and is used to clarify the severity of the potential hazard that warrants a warning. In this embodiment, the preset anomaly level is 4 or greater.

[0071] The above embodiment introduces a method for automatic identification of construction safety hazards based on deep learning from the perspective of method flow. The following embodiment introduces a system for automatic identification of construction safety hazards based on deep learning from the perspective of virtual modules or virtual units. For details, please see the following embodiment. The present application embodiment provides a construction safety hazard automatic identification system based on deep learning, such as Figure 2 As shown, the system 20 may specifically include: Information acquisition module 21, used to obtain personnel behavior information, equipment operation information, and environmental parameter information. Equipment operation information includes the operation information of each device itself and the interactive operation information of related devices. Environmental parameter information is the environmental information of the area where the device is located; The collection and extraction module 22 is used to collect historical construction abnormality information and the construction abnormality causes corresponding to each construction abnormality event in the historical construction abnormality information, and perform feature extraction on the historical construction abnormality information to obtain time period features and personnel behavior features, environmental parameter features, and equipment operation features corresponding to the time period features. The time period features include a first time period before each construction abnormality event occurs, a time period during the occurrence, and a second time period after the occurrence; The model training module 23 is used to train the preset construction abnormality hidden danger model based on the personnel behavior characteristics, environmental parameter characteristics and equipment operation characteristics to obtain the construction abnormality hidden danger model; The information processing module 24 is used to process the personnel behavior information, equipment operation information and environmental parameter information according to the personnel behavior characteristics, environmental parameter characteristics and equipment operation characteristics to obtain construction precursor data; The hidden danger identification module 25 is used to input the construction precursor data into the construction abnormal hidden danger model for identification and obtain the hidden danger identification result; The weight analysis module 26 is used to determine whether there is at least one hidden danger anomaly in the hidden danger identification results. If so, it determines the location of the area where the hidden danger anomaly exists and performs hidden danger weight analysis on the area location to obtain the hidden danger coefficient corresponding to the area location; The hidden danger determination module 27 is used to determine the hidden danger level based on the hidden danger coefficient corresponding to the regional location and the construction abnormality data and the cause of the construction abnormality in the hidden danger identification result to obtain the hidden danger abnormality level; The hidden danger warning module 28 is used to determine whether the hidden danger abnormality level meets the preset abnormality level. If it meets the level, a hidden danger abnormality warning is generated based on the construction abnormality data and the cause of the construction abnormality.

[0072] In one possible implementation, the model training module 23 trains a preset construction anomaly hidden danger model based on personnel behavior characteristics, environmental parameter characteristics, and equipment operation characteristics to obtain a construction anomaly hidden danger model, specifically for: determining a first behavior distinguishing feature, a first environmental parameter feature difference, and a first device operation feature difference based on the personnel behavior features, environmental parameter features, and device operation features corresponding to the second time period and the personnel behavior features, environmental parameter features, and device operation features corresponding to the first time period; Determine a second behavior distinguishing feature, a second environmental parameter feature difference, and a second equipment operation feature difference based on the personnel behavior features, environmental parameter features, and equipment operation features corresponding to the second time period and the personnel behavior features, environmental parameter features, and equipment operation features corresponding to the time period of the occurrence process; Sorting out construction abnormality trends based on the first behavioral distinguishing features, the first environmental parameter characteristic differences, and the first equipment operation characteristic differences to obtain precursor dynamic data corresponding to each construction abnormality event; Sorting out the construction anomaly based on the second behavior distinguishing feature, the second environmental parameter feature difference, and the second equipment operation feature difference to obtain construction anomaly data corresponding to each construction anomaly event; The precursor dynamic data, construction abnormality data and causes of construction abnormalities are input into the preset construction abnormality hidden danger model according to the time nodes for training to obtain the construction abnormality hidden danger model.

[0073] In one possible implementation, when the information processing module 24 processes the personnel behavior information, the equipment operation information, and the environmental parameter information according to the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain the construction precursor data, it is specifically used to: Extract features from personnel behavior information, equipment operation information, and environmental parameter information to obtain real-time behavior features, real-time operation features, and real-time parameter features; Performing feature matching on the real-time parameter features and the real-time operation features with the environmental parameter features and the equipment operation features during the time period of the occurrence process, respectively, to determine whether there is an operation feature match and / or parameter feature match; if not, performing feature matching on the real-time behavior features, the real-time operation features, and the real-time parameter features with the personnel behavior features, the environmental parameter features, and the equipment operation features during the second time period after the time graph period features occur, respectively, to obtain comparison behavior features, comparison parameter features, and comparison operation features; Based on the comparison of behavior features, parameter features and operation features, feature comparison processing is performed on real-time behavior features, real-time operation features and real-time parameter features to obtain construction precursor data.

[0074] In one possible implementation, when the information processing module 24 performs feature comparison processing on the real-time behavior features, the real-time operation features, and the real-time parameter features based on the comparison behavior features, the comparison parameter features, and the comparison operation features to obtain construction precursor data, it is specifically configured to: Perform behavioral feature differentiation analysis on the compared behavioral features and the real-time behavioral features to obtain real-time behavioral differentiation features; Perform parameter feature comparison on the real-time parameter feature and the comparison parameter feature to obtain the real-time environment parameter feature difference; Performing operation characteristic distinction comparison between the real-time operation characteristic and the comparison operation characteristic to obtain the real-time operation parameter characteristic difference; The trends of real-time behavioral distinguishing features, real-time environmental parameter characteristic differences, and real-time operating parameter characteristic differences are sorted out to obtain construction precursor data.

[0075] In a possible implementation, when the weight analysis module 26 performs a hidden danger weight analysis on a regional location to obtain a hidden danger coefficient corresponding to the regional location, it is specifically configured to: Acquire an image of the construction area and construct a three-dimensional construction model of the construction area based on the image of the construction area; Determine concentrated hazard areas that meet preset conditions and dispersed hazard areas that do not meet preset conditions based on the construction 3D model. The preset conditions are a specific number of people and adjacent equipment that meet a specific distance. The concentrated hidden danger areas and dispersed hidden danger areas are respectively regarded as hidden danger area points, and the hidden danger area points are connected according to the time node sequence corresponding to the abnormal construction events to obtain the hidden danger trend curve; Conduct regular extension analysis on the hidden danger trend curve to obtain the predicted trend curve of the probability of hidden danger anomalies occurring in the future period; Determine the future regional points corresponding to the predicted abnormal trend curve, and determine the hidden danger coefficients corresponding to different regional locations in the construction area based on the concentration and dispersion characteristics of the future regional points and the hidden danger area points.

[0076] In a possible implementation, when the weight analysis module 26 performs regular extension analysis on the hidden danger trend curve to obtain a predicted trend curve of the probability of hidden danger anomalies occurring in a future period, it is specifically used to: Determine the hidden danger development vector extending from different hidden danger area points to the next hidden danger area point according to the hidden danger trend curve, and the abnormal feature set corresponding to each hidden danger area point; Determine the hidden danger vector feature based on the hidden danger development vector, and add the hidden danger vector feature to the abnormal feature set to obtain the hidden danger feature set; Arrange each hidden danger feature set in a matrix according to the time sequence to obtain the hidden danger feature matrix corresponding to each hidden danger feature set; The hidden danger feature matrix is ​​deduced from the hidden danger feature set according to the time period to obtain multiple hidden danger point matrices in the future period; The future hidden danger feature set is determined based on multiple hidden danger point matrices, and the future hidden danger feature set is imported into the construction three-dimensional model to obtain a prediction trend curve of the probability of hidden danger anomalies occurring in the future period.

[0077] It can be understood that the specific working process of the above-described deep learning-based construction safety hazard automatic identification system 20 can refer to the implementation process of the corresponding steps in the aforementioned method embodiment, and will not be repeated here.

[0078] The present application also provides an electronic device, such as Figure 3 As shown, electronic device 300 includes: a processor 301 and a memory 303. Processor 301 and memory 303 are connected, for example, via a bus 302. Optionally, electronic device 300 may further include a transceiver 304. It should be noted that in actual applications, the number of transceivers 304 is not limited to one, and the structure of electronic device 300 does not constitute a limitation on the embodiments of this application.

[0079] Processor 301 can be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 301 can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0080] The bus 302 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 302 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 Only one thick line is used in the diagram, but it does not mean that there is only one bus or one type of bus.

[0081] The memory 303 may be a ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, a RAM (Random Access Memory) or other type of dynamic storage device that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0082] The memory 303 is used to store application code for executing the solution of the present application, and the execution is controlled by the processor 301. The processor 301 is used to execute the application code stored in the memory 303 to implement the content shown in the above method embodiment.

[0083] Among them, electronic devices include but are not limited to: mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The electronic device shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0084] An embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer-readable storage medium is run on a computer, the computer can execute the corresponding contents of the aforementioned method embodiment.

[0085] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown in sequence as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the flowcharts of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0086] Therefore, although the embodiments of the present invention have been described above, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the present invention, and all of these changes and modifications do not depart from the scope of protection of the present invention.

Claims

1. A method for automatically identifying construction safety hazards based on deep learning, characterized by: include: Obtaining personnel behavior information, equipment operation information, and environmental parameter information. The equipment operation information includes the operation information of each device itself and the interactive operation information of related devices. The environmental parameter information refers to the environmental information of the area where the device is located. Collecting historical construction abnormality information and the cause of each construction abnormality event in the historical construction abnormality information, and performing feature extraction on the historical construction abnormality information to obtain time period features and personnel behavior features, environmental parameter features, and equipment operation features corresponding to the time period features, wherein the time period features include a first time period before each construction abnormality event occurs, a time period during the occurrence, and a second time period after the occurrence; Training a preset construction abnormality hidden danger model based on the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain a construction abnormality hidden danger model; Processing the personnel behavior information, the equipment operation information, and the environmental parameter information according to the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain construction precursor data; Inputting the construction precursor data into the construction abnormality hidden danger model for identification to obtain hidden danger identification results; Determining whether there is at least one hidden danger anomaly in the hidden danger identification result; if so, determining the location of the region where the hidden danger anomaly exists, and performing a hidden danger weight analysis on the region location to obtain a hidden danger coefficient corresponding to the region location; Determine the hidden danger level based on the hidden danger coefficient corresponding to the regional location and the construction abnormality data and the cause of the construction abnormality in the hidden danger identification result to obtain the hidden danger abnormality level; Determine whether the hidden danger abnormality level meets the preset abnormality level. If so, generate a hidden danger abnormality warning based on the construction abnormality data and the cause of the construction abnormality.

2. The method for automatically identifying construction safety hazards based on deep learning according to claim 1, characterized in that: The training of a preset construction abnormality hidden danger model based on the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain the construction abnormality hidden danger model includes: determining a first behavior distinguishing feature, a first environmental parameter feature difference, and a first device operation feature difference based on the personnel behavior features, environmental parameter features, and device operation features corresponding to the second time period and the personnel behavior features, environmental parameter features, and device operation features corresponding to the first time period; Determine a second behavior distinguishing feature, a second environmental parameter feature difference, and a second device operation feature difference based on the personnel behavior features, environmental parameter features, and device operation features corresponding to the second time period and the personnel behavior features, environmental parameter features, and device operation features corresponding to the time period of the occurrence process; Sorting out construction abnormality trends based on the first behavioral distinguishing features, the first environmental parameter characteristic differences, and the first equipment operation characteristic differences to obtain precursor dynamic data corresponding to each construction abnormality event; Sorting out construction anomalies based on the second behavior distinguishing feature, the second environmental parameter feature difference, and the second equipment operation feature difference to obtain construction anomaly data corresponding to each construction anomaly event; The precursor dynamic data, the construction abnormality data and the cause of the construction abnormality are respectively input into a preset construction abnormality hidden danger model according to time nodes for training to obtain a construction abnormality hidden danger model.

3. The method for automatically identifying construction safety hazards based on deep learning according to claim 2, characterized in that: The personnel behavior information, the equipment operation information, and the environmental parameter information are processed according to the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain construction precursor data, including: Extracting features from the personnel behavior information, the equipment operation information, and the environmental parameter information to obtain real-time behavior features, real-time operation features, and real-time parameter features; Performing feature matching on the real-time parameter feature and the real-time operation feature with the environmental parameter feature and the equipment operation feature during the time period during which the time period feature occurs, respectively, to determine whether there is an operation feature match and / or a parameter feature match; if not, performing feature matching on the real-time behavior feature, the real-time operation feature, and the real-time parameter feature with the human behavior feature, the environmental parameter feature, and the equipment operation feature during a second time period after the time period feature occurs, respectively, to obtain a comparison behavior feature, a comparison parameter feature, and a comparison operation feature; Based on the comparison behavior feature, the comparison parameter feature and the comparison operation feature, feature comparison processing is performed on the real-time behavior feature, the real-time operation feature and the real-time parameter feature to obtain construction precursor data.

4. The method for automatically identifying construction safety hazards based on deep learning according to claim 3, characterized in that: The performing feature comparison processing on the real-time behavior feature, the real-time operation feature, and the real-time parameter feature based on the comparison behavior feature, the comparison parameter feature, and the comparison operation feature to obtain construction precursor data includes: Performing behavioral feature distinction analysis on the compared behavioral features and the real-time behavioral features to obtain real-time behavioral distinction features; Perform parameter feature difference comparison on the real-time parameter feature and the comparison parameter feature to obtain a real-time environment parameter feature difference; Performing operation characteristic difference comparison between the real-time operation characteristic and the comparison operation characteristic to obtain a real-time operation parameter characteristic difference; The real-time behavior distinguishing features, the real-time environmental parameter characteristic differences, and the real-time operating parameter characteristic differences are trended to obtain construction precursor data.

5. The method for automatically identifying construction safety hazards based on deep learning according to claim 1, characterized in that: The performing of hidden danger weight analysis on the regional location to obtain the hidden danger coefficient corresponding to the regional location includes: Acquire an image of the construction area, and construct a three-dimensional construction model of the construction area based on the image of the construction area; Determining, based on the three-dimensional construction model, concentrated hazard areas that meet preset conditions and dispersed hazard areas that do not meet preset conditions, wherein the preset conditions are a set number of personnel and adjacent equipment that meet a set distance; The concentrated hidden danger area and the dispersed hidden danger area are respectively used as hidden danger area points, and the hidden danger area points are connected according to the time node sequence corresponding to the abnormal construction event to obtain a hidden danger trend curve; Perform regular extension analysis on the hidden danger trend curve to obtain a predicted trend curve of the probability of hidden danger anomalies occurring in future periods; Determine the future regional points corresponding to the predicted abnormal trend curve, and determine the hidden danger coefficients corresponding to different regional positions in the construction area based on the concentration and dispersion characteristics of the future regional points and the hidden danger regional points.

6. The method for automatically identifying construction safety hazards based on deep learning according to claim 5, characterized in that: The regular extension analysis of the hidden danger trend curve is performed to obtain a prediction trend curve of the probability of hidden danger anomalies occurring in future periods, including: Determine, based on the hidden danger trend curve, the hidden danger development vector extending from different hidden danger area points to the next hidden danger area point and the abnormal feature set corresponding to each hidden danger area point; Determining a hidden danger vector feature based on the hidden danger development vector, and adding the hidden danger vector feature to the abnormal feature set to obtain a hidden danger feature set; Arranging each of the hidden danger feature sets in a matrix according to a time sequence to obtain a hidden danger feature matrix corresponding to each hidden danger feature set; Deducing the hidden danger feature matrix from the hidden danger feature set according to the time period to obtain multiple hidden danger point matrices in the future period; A future hidden danger feature set is determined based on multiple hidden danger point matrices, and the future hidden danger feature set is imported into the construction three-dimensional model to obtain a prediction trend curve of the probability of hidden danger anomalies occurring in future periods.

7. The automatic identification system for construction safety hazards based on deep learning is characterized by: include: An information acquisition module is used to obtain personnel behavior information, equipment operation information, and environmental parameter information. The equipment operation information includes the operation information of each device itself and the interactive operation information of related devices. The environmental parameter information is the environmental information of the area where the device is located; a collection and extraction module for collecting historical construction abnormality information and the cause of each construction abnormality event in the historical construction abnormality information, and performing feature extraction on the historical construction abnormality information to obtain time period features and personnel behavior features, environmental parameter features, and equipment operation features corresponding to the time period features, wherein the time period features include a first time period before each construction abnormality event occurs, a time period during the occurrence, and a second time period after the occurrence; A model training module is used to train a preset construction abnormality hidden danger model based on the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain a construction abnormality hidden danger model; an information processing module, configured to process the personnel behavior information, the equipment operation information, and the environmental parameter information according to the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain construction precursor data; a hidden danger identification module, configured to input the construction precursor data into the construction abnormal hidden danger model for identification, and obtain hidden danger identification results; a weight analysis module, configured to determine whether there is at least one hidden danger anomaly in the hidden danger identification result; if so, determine the location of the region where the hidden danger anomaly exists, and perform hidden danger weight analysis on the region to obtain a hidden danger coefficient corresponding to the region; a hidden danger determination module, configured to determine the hidden danger level based on the hidden danger coefficient corresponding to the regional location and the construction abnormality data and the cause of the construction abnormality in the hidden danger identification result, and obtain a hidden danger abnormality level; The hidden danger warning module is used to determine whether the hidden danger abnormality level meets the preset abnormality level. If it does, it generates a hidden danger abnormality warning based on the construction abnormality data and the cause of the construction abnormality.

8. The deep learning-based automatic identification system for construction safety hazards according to claim 7, characterized in that: The model training module is specifically used to train the preset construction abnormality hidden danger model based on the personnel behavior characteristics, the environmental parameter characteristics, and the equipment operation characteristics to obtain the construction abnormality hidden danger model: determining a first behavior distinguishing feature, a first environmental parameter feature difference, and a first device operation feature difference based on the personnel behavior features, environmental parameter features, and device operation features corresponding to the second time period and the personnel behavior features, environmental parameter features, and device operation features corresponding to the first time period; Determine a second behavior distinguishing feature, a second environmental parameter feature difference, and a second device operation feature difference based on the personnel behavior features, environmental parameter features, and device operation features corresponding to the second time period and the personnel behavior features, environmental parameter features, and device operation features corresponding to the time period of the occurrence process; Sorting out construction abnormality trends based on the first behavioral distinguishing features, the first environmental parameter characteristic differences, and the first equipment operation characteristic differences to obtain precursor dynamic data corresponding to each construction abnormality event; Sorting out construction anomalies based on the second behavior distinguishing feature, the second environmental parameter feature difference, and the second equipment operation feature difference to obtain construction anomaly data corresponding to each construction anomaly event; The precursor dynamic data, the construction abnormality data and the cause of the construction abnormality are respectively input into a preset construction abnormality hidden danger model according to time nodes for training to obtain a construction abnormality hidden danger model.

9. An electronic device comprising a memory and a processor; the memory stores a computer program; and When the processor executes the computer program, the method for automatically identifying construction safety hazards based on deep learning as described in any one of claims 1 to 6 is 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 method for automatically identifying construction safety hazards based on deep learning as described in any one of claims 1 to 6 is implemented.

Images