Iron tower safety operation monitoring method and system based on intelligent AI recognition, and medium

CN121581620BActive Publication Date: 2026-09-15TIETA ZHILIAN HEBEI CO LTD +3
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202511502065.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-10-21
Publication Date
2026-09-15
Estimated Expiration
2045-10-21

AI Technical Summary

Technical Problem

[0003]本申请提供了智能AI识别的铁塔安全作业监测方法、系统及介质,旨在解决现有技术的铁塔安全作业监测依赖于人工巡检与传统传感器监测,导致对作业现场的实时监控能力不足,无法实时及时识别安全隐患和异常状况,从而增加事故发生概率的技术问题

Benefits of technology

[0008]By collecting data from the tower operation site, a multidimensional dataset is obtained. Based on the type of acquisition equipment, the data is hierarchically managed, enabling precise acquisition of various operation-related information, avoiding information redundancy, and improving monitoring accuracy and efficiency. The data is layered into multiple data channels, each focusing on different types of information, making data analysis more targeted and efficient. The first data channel detects and locks onto targets at the tower operation site, capturing personnel and objects in real time during the operation, ensuring accurate identification of operational behaviors. The second data channel identifies personnel behavior and determines operational behavior parameters, identifying specific behaviors in different operational scenarios, providing a reliable basis for subsequent safety assessments. The third data channel integrates environmental data. By analyzing and combining operational behavior parameters for safety assessment, the system can comprehensively evaluate the safety situation at the work site. Utilizing AI technology for intelligent identification and dynamic monitoring, it can update the risk level of the work site in real time, ensuring timely response to any changes in the environment and behavior, thereby improving operational safety. Based on the assessment results, a dynamic risk level is generated, and safety alarms are issued for the work site based on this level. Alarm information is pushed to remote terminals for risk management, ensuring that risk information reaches the responsible parties quickly and enabling timely response. The remote terminal takes corresponding control measures based on the risk level, effectively preventing potential safety accidents. Through real-time monitoring and alarm mechanisms, the safety of tower operations is significantly improved, the probability of accidents is reduced, and the lives of workers are protected.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121581620B_ABST
    Figure CN121581620B_ABST
Patent Text Reader

Abstract

The application provides a tower safety operation monitoring method and system based on intelligent AI recognition, and a medium, and relates to the technical field of intelligent monitoring.The method comprises the following steps: collecting data of a tower operation site, obtaining a multidimensional data set, performing data layering, and determining a plurality of data channels; performing target detection through a first data channel, performing personnel behavior recognition according to the locked target in combination with a second data channel, and determining operation behavior parameters; performing environment fusion analysis through a third data channel, performing safety evaluation, generating safety situation quantitative parameters for intelligent AI risk dynamic monitoring and recognition, and generating a dynamic risk level; and generating alarm information and pushing the alarm information to a remote terminal for risk control.The application solves the technical problem that the tower safety operation monitoring of the prior art relies on manual inspection and traditional sensor monitoring, which leads to insufficient real-time monitoring capability of the operation site, the inability to identify safety hazards and abnormal conditions in real time, and thus increases the probability of accidents.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology, specifically to a method, system, and medium for monitoring safe operation of iron towers using intelligent AI recognition. Background Technology

[0002] As a vital infrastructure for industries such as communications, broadcasting, and power, towers play a crucial role. The construction, maintenance, and repair of towers are typically carried out in high-altitude, complex, and hazardous environments, making safety monitoring paramount. Current technologies rely on manual inspections or basic sensors for safety monitoring of tower operations. Due to the dangerous and complex working environment, manual inspections face challenges such as workers operating at heights and long work cycles, making it difficult to effectively detect safety hazards and anomalies in real time. This leads to potential safety risks going unidentified, increasing the probability of accidents. Traditional sensors can only detect environmental parameters such as wind speed and temperature, failing to accurately capture personnel behavior and equipment status at the work site, and are unable to conduct comprehensive risk assessments. This results in the inability to promptly detect various safety hazards in complex working environments, such as whether personnel are in dangerous positions or whether equipment is malfunctioning, thus increasing the risk of accidents to workers. Summary of the Invention

[0003] This application provides a method, system, and medium for monitoring safe operation of iron towers using intelligent AI recognition. It aims to solve the technical problem that existing iron tower safety operation monitoring relies on manual inspections and traditional sensor monitoring, resulting in insufficient real-time monitoring capabilities of the work site and the inability to identify safety hazards and abnormal conditions in a timely manner, thereby increasing the probability of accidents.

[0004] The first aspect disclosed in this application provides a method for monitoring safe operations on iron towers using intelligent AI recognition. The method includes: collecting data from the iron tower operation site to obtain a multidimensional dataset; layering the multidimensional dataset to determine multiple data channels, where each data channel corresponds to a data layer, and the multiple data channels include a first data channel, a second data channel, and a third data channel; performing target detection through the first data channel; identifying personnel behavior based on the locked target and the second data channel to determine operational behavior parameters; performing environmental fusion analysis through the third data channel; conducting a safety assessment based on the environmental fusion feature set and the operational behavior parameters; generating quantitative parameters for safety status; performing intelligent AI-based dynamic risk monitoring and identification; and generating a dynamic risk level; issuing a safety alarm at the iron tower operation site based on the dynamic risk level; and pushing alarm information to a remote terminal for risk management.

[0005] The second aspect of this application discloses an intelligent AI-based tower safety operation monitoring system. This system is used in the aforementioned intelligent AI-based tower safety operation monitoring method. The system includes: a data acquisition module for collecting data from the tower operation site to obtain a multidimensional dataset; layering the multidimensional dataset to determine multiple data channels, where each data channel corresponds to a data layer and includes a first data channel, a second data channel, and a third data channel; a personnel behavior recognition module for target detection via the first data channel, and personnel behavior recognition based on the locked target and the second data channel to determine operation behavior parameters; a safety assessment module for environmental fusion analysis via the third data channel, and safety assessment based on the environmental fusion feature set and the operation behavior parameters to generate quantitative safety status parameters for intelligent AI risk dynamic monitoring and identification, generating a dynamic risk level; and a safety alarm module for issuing safety alarms at the tower operation site based on the dynamic risk level, generating alarm information and pushing it to a remote terminal for risk management.

[0006] The third aspect disclosed in this application provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the intelligent AI-based tower safety operation monitoring method in the first aspect.

[0007] One or more technical solutions provided in this application have at least the following beneficial effects:

[0008] By collecting data from the tower operation site, a multidimensional dataset is obtained. Based on the type of acquisition equipment, the data is hierarchically managed, enabling precise acquisition of various operation-related information, avoiding information redundancy, and improving monitoring accuracy and efficiency. The data is layered into multiple data channels, each focusing on different types of information, making data analysis more targeted and efficient. The first data channel detects and locks onto targets at the tower operation site, capturing personnel and objects in real time during the operation, ensuring accurate identification of operational behaviors. The second data channel identifies personnel behavior and determines operational behavior parameters, identifying specific behaviors in different operational scenarios, providing a reliable basis for subsequent safety assessments. The third data channel integrates environmental data. By analyzing and combining operational behavior parameters for safety assessment, the system can comprehensively evaluate the safety situation at the work site. Utilizing AI technology for intelligent identification and dynamic monitoring, it can update the risk level of the work site in real time, ensuring timely response to any changes in the environment and behavior, thereby improving operational safety. Based on the assessment results, a dynamic risk level is generated, and safety alarms are issued for the work site based on this level. Alarm information is pushed to remote terminals for risk management, ensuring that risk information reaches the responsible parties quickly and enabling timely response. The remote terminal takes corresponding control measures based on the risk level, effectively preventing potential safety accidents. Through real-time monitoring and alarm mechanisms, the safety of tower operations is significantly improved, the probability of accidents is reduced, and the lives of workers are protected.

[0009] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0010] Figure 1 A schematic diagram of the intelligent AI-based tower safety operation monitoring method provided in this application embodiment.

[0011] Figure 2 A schematic diagram of the structure of the intelligent AI recognition tower safety operation monitoring system provided in this application embodiment.

[0012] Figure labeling: Data acquisition module 10, personnel behavior recognition module 20, security assessment module 30, security alarm module 40. Detailed Implementation

[0013] This application provides a method, system, and medium for monitoring safe operations on iron towers using intelligent AI recognition. It solves the technical problem that existing methods for monitoring safe operations on iron towers rely on manual inspections and traditional sensor monitoring, resulting in insufficient real-time monitoring capabilities of the work site and an inability to identify safety hazards and abnormal conditions in a timely manner, thereby increasing the probability of accidents.

[0014] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0015] Example 1, as Figure 1 As shown in the embodiment of this application, a method for monitoring safe operation of iron towers using intelligent AI recognition is provided. The method includes:

[0016] Data is collected from the tower operation site to obtain a multidimensional dataset. The multidimensional dataset is then layered to determine multiple data channels. These multiple data channels correspond to the data layers and include a first data channel, a second data channel, and a third data channel.

[0017] At the tower operation site, a variety of sensors are used to collect data. These sensors collect different types of information to form a multidimensional dataset. A multidimensional dataset means that there is sensor data from multiple dimensions at the same point in time, such as image data, environmental monitoring data (such as wind speed, temperature, electromagnetic interference, etc.), and personnel behavior data (such as posture, movement, etc.).

[0018] The collected multidimensional dataset is processed hierarchically. This process divides different types of data into multiple data layers according to different levels, including image acquisition data layer, behavior perception data layer, and environmental monitoring data layer. Each data layer is dedicated to processing a specific type of data. Based on these data layers, multiple data channels are determined to process different types of data. Each data channel is connected to the corresponding data layer. For example, the first data channel is used to process image data for target detection; the second data channel is used to process behavior perception data for personnel behavior recognition; and the third data channel processes environmental data for environmental fusion analysis.

[0019] Target detection is performed through the first data channel, and personnel behavior is identified by combining the locked target with the second data channel to determine the operation behavior parameters.

[0020] The first data channel uses image data obtained from the image acquisition data layer for target detection. This process uses AI algorithms (such as convolutional neural networks) to identify target objects such as tower workers, equipment, and tools in video frames. In target detection, all relevant targets are identified, and their location coordinates are marked on the image in the form of bounding boxes. The second data channel uses personnel action and posture data from the behavior perception data layer to identify personnel behavior. Here, the actions of each target are analyzed, such as climbing the tower or using tools. The personnel behavior recognition process is based on the target's posture (such as skeletal structure) and temporal data to identify personnel behavior patterns. After locking the target and performing personnel behavior recognition, the target's work behavior is further analyzed to generate relevant work behavior parameters, including the type of action of the worker, the work status, and the adaptability to the work environment, such as whether they are in a dangerous work area or whether they are using appropriate protective equipment.

[0021] Environmental fusion analysis is performed through the third data channel. Based on the environmental fusion feature set and the operational behavior parameters, a safety assessment is conducted to generate quantitative parameters of the safety situation. Intelligent AI risk dynamic monitoring and identification are then performed to generate dynamic risk levels.

[0022] Environmental fusion analysis refers to the weighted and combined analysis of multiple environmental parameters to form a comprehensive environmental fusion feature set. This process involves analyzing multiple environmental parameters such as wind speed, temperature, electromagnetic interference, and vibration to construct a feature set reflecting the environment at the tower operation site. For example, excessively high wind speeds can affect personnel safety, and electromagnetic interference can affect the operation of equipment or instruments by workers. By analyzing the relationship between the environmental fusion feature set and operational behavior parameters, a safety assessment is conducted. This process uses multi-dimensional data correlation analysis to combine environmental parameters with personnel behavior, assess the overall safety situation at the site, and generate quantitative safety situation parameters, including wind speed safety situation parameters, vibration safety situation parameters, and electromagnetic safety situation parameters.

[0023] Based on the generated quantitative parameters of the safety situation, intelligent AI risk dynamic monitoring and identification are carried out. The AI ​​model performs predictive analysis based on historical data, environmental characteristics, and work behavior, monitors the risk changes on site in real time, and generates a dynamic risk level for each risk factor through multi-dimensional fusion analysis. The dynamic risk level reflects the degree of impact of each parameter on safety during real-time operation.

[0024] Based on the dynamic risk level, safety alarms are generated at the tower operation site, and alarm information is pushed to remote terminals for risk management.

[0025] Based on the generated dynamic risk level, the overall safety of the work site is assessed. When the dynamic risk level reaches the set safety threshold, a safety alarm is immediately triggered. Key risk core fields, such as specific hazard factors, workers, and environmental parameters, are extracted from the risk assessment and packaged into a risk data packet for subsequent processing and response. The risk data packet is transmitted over the network and pushed to remote terminals, including managers' mobile phones, computers, or monitoring systems. After the packet is pushed, safety management personnel take corresponding countermeasures based on the risk information. For example, for high-altitude operations in excessively high wind speeds, the operation is suspended and a safety instruction is issued.

[0026] Furthermore, data is collected from the tower operation site to obtain a multidimensional dataset. This multidimensional dataset is then layered to determine multiple data channels. These multiple data channels correspond to data layers and include a first data channel, a second data channel, and a third data channel. The method includes:

[0027] Based on the multidimensional dataset, the data is layered according to the type of acquisition device to obtain an image acquisition data layer, a behavior perception data layer, and an environmental monitoring data layer. The image acquisition data layer transmits video frame sequences for compression and allocation to construct the first data channel. The behavior perception data layer transmits personnel posture data for calculation to construct the second data channel. The environmental monitoring data layer transmits environmental physical parameters for filtering to construct the third data channel.

[0028] The multidimensional dataset contains data from different types of sensors, such as image sensors, behavior sensing sensors, and environmental monitoring sensors. These sensors collect data with different characteristics; therefore, the data is layered according to the type of acquisition device to ensure that each layer can be analyzed and processed independently. Specifically, the image acquisition data layer comes from video cameras, including video frame sequences such as high-resolution images and video streams, used for tasks such as target detection, image analysis, and object recognition. The behavior sensing data layer comes from depth cameras, motion sensors, etc., used to capture personnel posture and behavior data, focusing on personnel movements, postures, and dynamics, supporting the recognition and analysis of work behaviors. The environmental monitoring data layer includes data collected by environmental sensors, such as wind speed, temperature, humidity, vibration, and electromagnetic interference, used for environmental safety monitoring and risk assessment.

[0029] The video frame sequence is acquired from on-site camera equipment. It consists of multiple consecutive video frames containing dynamic changes in the scene, used to monitor the real-time situation on-site. Due to the large amount of video data, transmission and storage may be limited by bandwidth and storage constraints. Therefore, the video frame sequence needs to be compressed. Compression can reduce the amount of data, increase transmission speed, and reduce the consumption of computing resources. Common compression technologies include video coding standards such as H.264 and H.265. After compression, the video frame sequence forms the first data channel for transmitting and processing image data to support functions such as target detection, personnel recognition, and on-site monitoring.

[0030] Personnel posture data includes the movement trajectory and posture changes of the human body. Posture data processing refers to inferring information such as the position of a person's limbs and movement patterns by analyzing sensor data. For example, by capturing the skeletal structure of a person through a depth camera, the spatial coordinates of each joint can be calculated, and the person's behavior, such as climbing a tower, standing, or carrying tools, can be inferred. After processing, a second data channel is constructed for the identification and analysis of personnel behavior. By analyzing this data, the work status of personnel can be assessed, potential hazardous behaviors can be identified, and work behavior parameters can be generated, such as whether there are dangerous actions or non-standard operations.

[0031] Environmental physical parameters reflect the on-site environmental conditions. Since environmental sensors may be affected by noise or errors, the transmitted data needs to be filtered. Filtering removes inaccurate or irrelevant data to ensure high accuracy. Common filtering methods include Kalman filtering and mean filtering. After filtering, a third data channel is constructed for analyzing environmental parameters to support subsequent safety assessments and risk identification.

[0032] Furthermore, the method involves target detection via the first data channel, followed by personnel behavior recognition based on the locked target and the second data channel, to determine operational behavior parameters.

[0033] Based on the first data channel, video frame analysis is performed on the tower operation site according to the video frame sequence to obtain a video data stream; multiple target detection information of the tower operation site is retrieved, and based on the video data stream, the multiple target detection information is used for identification to obtain a locked target, the locked target including target bounding box coordinates; spatial positioning is performed according to the target bounding box coordinates to determine personnel spatial location information, and the personnel spatial location information is synchronized to the second data channel for personnel behavior recognition to determine the operation behavior parameters.

[0034] A video frame sequence consists of multiple consecutive video frames that contain dynamic information about the scene, such as the background, tower structure, and workers. These video frames are preprocessed, such as denoising and contrast adjustment, to ensure the accuracy of subsequent processing. The processed video data stream is then used for subsequent target detection and behavior recognition tasks.

[0035] Target detection information refers to the feature data of multiple known target categories at the work site, such as workers and equipment. Pre-trained target detection models are retrieved. These models have learned how to distinguish different categories of targets through target detection information. Each frame in the video data stream is processed by the target detection model to identify and mark the locked target, surround it in a rectangular box, and return the coordinates of the box to obtain the target bounding box coordinates. The target bounding box coordinates are usually the coordinates of the upper left and lower right corners of a rectangle, representing the position of the locked target in the image.

[0036] Spatial positioning refers to converting a locked target from a two-dimensional image space into three-dimensional world coordinates based on the target bounding box coordinates in a video frame. Using the target bounding box coordinates and the camera's internal and external parameters, such as focal length and viewing angle, the target's position in physical space can be calculated. For example, in the case of a single camera, deep learning methods can be used to estimate the distance between the object and the camera, thereby performing spatial positioning and determining personnel spatial location information, such as whether their location is in a high-risk area of ​​a tower or near dangerous equipment. This personnel spatial location information is synchronized to a second data channel. Through this second data channel, a human pose estimation algorithm is used to analyze the specific actions and behaviors of personnel based on their skeletal structure in the video frame. Based on the pose data and spatial location information, further identification of worker behaviors, such as climbing towers, carrying heavy objects, and using tools, is achieved, and work behavior parameters are generated based on these behaviors.

[0037] Furthermore, spatial positioning is performed based on the target bounding box coordinates to determine the personnel's spatial location information. This personnel spatial location information is then synchronized to the second data channel for personnel behavior recognition to determine the work behavior parameters. The method includes:

[0038] The second data channel is used to identify personnel at the tower operation site according to the operation sequence, and obtain personnel skeletal joint time sequence data; based on the personnel skeletal joint time sequence data, time sequence is extracted to determine multiple time window segments; according to the multiple time window segments, the personnel spatial location information is synchronized to the second data channel and associated with the personnel skeletal joint time sequence data to determine the operation behavior parameters.

[0039] Operation timing refers to the organization and recording of activities at the tower operation site in chronological order. Each operation action has a time tag, which enables the identification and tracking of personnel based on the order of operations. Through the second data channel, personnel identification information at the operation site is obtained from the behavior perception data layer. This information mainly comes from the posture estimation of personnel. The movement of each human body is composed of the movement of multiple skeletal joints. By identifying the skeletal structure of personnel in each video frame, such as the positions of the head, shoulders, arms, and legs, the spatial coordinates of each joint are generated. These spatial coordinates are arranged in chronological order to form skeletal joint timing data. That is, throughout the entire operation, the spatial position of each skeletal joint is identified and recorded, and this data is synchronized with the time sequence of video frames.

[0040] Since the work process is a continuous action, single time-series data often contains too much information and is difficult to analyze effectively. Therefore, these long-term skeletal joint time-series data are segmented into multiple time windows for analysis. These time windows are divided based on the start and end times of the action. For example, when a person begins climbing a tower, the start time of that action can be defined as the beginning of a time window, and a time segment can be extracted at the end of the action. Each time window segment contains a continuous segment of skeletal joint data, which describes the specific actions of the worker within a specific time period. The formation of multiple time window segments allows the system to perform fine-grained analysis of the worker's behavior, no longer relying solely on the summary data of the entire work process, but analyzing each small segment of the action.

[0041] By matching the spatial location information of personnel in each time window with the time-series data of their skeletal joints, we can not only determine the personnel's movement trajectory but also the relationship between their position and specific actions within each time window. For example, if a person is climbing a tower during a certain time period, the spatial location information indicates that they are approaching the top of the tower. This information can be used to analyze whether there are any risks associated with working at height. Through the correlation and matching of time-series data, we can ultimately determine the parameters of the work behavior, including action type, action standardization, action duration, and identification of hazardous behaviors.

[0042] Furthermore, the method for determining the operational behavior parameters by synchronizing the personnel's spatial location information to the second data channel according to the multiple time window segments and associating and matching it with the personnel's skeletal joint time-series data includes:

[0043] The center coordinates of the target bounding box are extracted. Using these center coordinates as the center, the temporal data of the human skeletal joints are normalized according to the multiple time window segments to determine the three-dimensional coordinate sequence. Graph convolutional feature analysis is performed based on the three-dimensional coordinate sequence to extract spatial topological features. Specifically, the graph convolutional feature analysis uses a graph convolutional network (GCN), which contains three graph convolutional layers, each with 64 feature channels and a 3×3 kernel size. An adjacency matrix is ​​used to represent the connections between human skeletal joints, and neighbor node features are aggregated through a message passing mechanism. The ReLU function is used for activation. Temporal convolutional feature analysis is also performed based on the three-dimensional coordinate sequence to extract action evolution features. This temporal convolutional feature analysis uses a one-dimensional convolutional neural network (1D-CNN), which contains four convolutional layers with kernel sizes of [9, 5, 3, 3] and feature channel numbers of [64, 128, 256].

[256] , the step size is 1, the filling method is same, the pooling layer adopts max pooling, and the pooling window size is 2; according to the spatial topology features, the spatial location information of the personnel is associated and matched with the temporal data of the personnel's skeletal joints to obtain the first operation behavior parameter; according to the action evolution features, the spatial location information of the personnel is associated and matched with the temporal data of the personnel's skeletal joints to obtain the second operation behavior parameter.

[0044] The center coordinates of the target bounding box are extracted. The center point is the geometric center of the target bounding box, which is the average of the coordinates of the top-left and bottom-right corners of the rectangle. Normalization adjusts the data to a unified standard to eliminate the influence of different data scales or ranges. In this step, the three-dimensional coordinates of the human skeleton joint time series data are normalized based on the center coordinates. By standardizing the coordinates of each joint relative to the center coordinates, relative coordinates are obtained. This method helps the system remove scene variation factors during analysis, enabling unified analysis regardless of the target's absolute position and avoiding data bias caused by changes in position or scale. After normalization, a three-dimensional coordinate sequence is obtained, which describes the position of the human skeleton joints in three-dimensional space within each time window segment.

[0045] Graph convolution is a method for processing graph-structured data. It effectively captures the relationships between nodes in a graph. For human skeletal joints, each joint can be considered a node, while the connections between joints (such as the connection between the arm, leg, and torso) are considered edges. Graph convolution analysis can utilize the spatial topology of skeletal joints to extract the relative relationships between joints and their spatial distribution patterns. In a three-dimensional coordinate sequence, the spatial position of each skeletal joint is relative. Through graph convolution operations, the connections and relative positions between skeletal joints can be analyzed to extract spatial topological features, namely, the distribution of joints in space and their interrelationships. For example, by evaluating the relative positional relationships between the hands and shoulders, and the legs and waist, different postures and movement patterns can be identified.

[0046] Temporal convolution is a technique specifically designed for processing time-series data. It extracts time-series features by sliding a data window along the time dimension through convolution operations. Action evolution features refer to the changes in human actions over time, analyzing how actions transition from one state to another over time. For example, during a person's climbing action, the body's joints constantly adjust from start to finish, and these adjustments follow a predictable pattern. Through temporal convolution, dynamic features in this time series, such as movement speed, acceleration, and the duration of the action, are extracted to generate action evolution features. For example, whether a certain action is smooth, whether there is an acceleration or deceleration process, or whether there are sudden changes. By comparing the action evolution features in different time windows, it is possible to determine whether the worker's movement conforms to the specifications and whether there are any abnormal or unsafe actions.

[0047] Based on spatial topological features, the spatial relationships of skeletal joints are combined with the specific location of the personnel. For example, the relative position of a certain joint to other joints will affect the posture of the personnel. This information can be combined with the specific location of the personnel to determine their working status. Through matching, the dynamic posture and position distribution of personnel in different work activities can be identified, and their work behavior status can be inferred based on this information to obtain the first work behavior parameter. The first work behavior parameter represents the static behavior characteristics of the personnel in the work, which can help determine whether the current work status of the personnel is normal and whether there are potential safety risks.

[0048] Based on the characteristics of motion evolution, the spatial relationship of skeletal joints is combined with the specific position of the personnel to analyze the relationship between the change process of personnel's movements and their spatial movement trajectory. For example, when the personnel's movements change, do their spatial positions change accordingly, or do they experience acceleration or irregular movements? Through time series analysis, it is determined how personnel transition from one movement to another during the operation, such as from standing to climbing, or from normal operation to abnormal movement. The resulting second operational behavior parameter describes the dynamic behavioral characteristics of personnel in the operation, which is used to judge the work mode and efficiency of personnel in the operation and can provide support for further risk assessment.

[0049] Furthermore, environmental fusion analysis is performed through the third data channel. Based on the environmental fusion feature set and the operational behavior parameters, a safety assessment is conducted to generate quantitative parameters of the safety situation. This is then used for intelligent AI-based dynamic risk monitoring and identification, generating a dynamic risk level. The method includes:

[0050] The environmental physical parameters are analyzed to generate wind speed influence factors, vibration hazard coefficients, and electromagnetic interference levels. These factors are then fused and analyzed via the third data channel to construct an environmental fusion feature set. The fusion analysis employs a weighted fusion algorithm, multiplying each factor by its corresponding fusion weight coefficient and summing the results to obtain the environmental fusion feature set, where the sum of the three weight coefficients equals one. A multidimensional correlation analysis is then performed between the environmental fusion feature set and the operational behavior parameters to construct an environmental behavior correlation matrix. This matrix is ​​then iterated through according to the wind speed influence factor, vibration hazard coefficient, electromagnetic interference level, etc. A safety assessment is conducted at each level to generate quantitative parameters for wind speed safety status, vibration safety status, and electromagnetic safety status. Based on these parameters, safety risk weights are allocated, and multiple weight coefficients are determined. Intelligent AI-based dynamic risk monitoring and identification are performed on the wind speed safety status parameters according to these weight coefficients to generate a dynamic wind speed risk level. Similarly, intelligent AI-based dynamic risk monitoring and identification are performed on the vibration safety status parameters according to these weight coefficients to generate a dynamic vibration risk level. Finally, intelligent AI-based dynamic risk monitoring and identification are performed on the electromagnetic safety status parameters according to these weight coefficients to generate a dynamic electromagnetic risk level.

[0051] Environmental physical parameters refer to various external environmental factors at the tower operation site, such as wind speed, vibration, and electromagnetic interference. These directly affect the safety of workers. Among them, the wind speed impact factor refers to the degree of wind speed's influence on the working environment, especially for high-altitude operations. For example, strong winds may cause instability in the tower or loss of balance for workers, increasing the risk of accidents. The feature analysis process calculates the wind speed variation trend and extreme values ​​based on on-site wind speed data and quantifies them into a wind speed impact factor. This factor reflects the degree of risk of the working environment under different wind speed conditions. During operation, the tower and its surrounding equipment may experience... Vibration, if it exceeds a certain threshold, can cause workers to lose their balance or equipment to malfunction. Vibration monitoring equipment collects data, performs feature analysis, and calculates the vibration hazard coefficient. This coefficient reflects the intensity and frequency of vibration at the work site; the higher the coefficient, the greater the vibration hazard. Electromagnetic interference level reflects the potential impact of electromagnetic waves present at the site on workers or equipment. For example, electromagnetic interference may affect personnel's communication equipment or electronic equipment on towers, or even cause instability in the work environment. Electromagnetic monitoring equipment collects data, analyzes information such as the intensity and frequency of electromagnetic interference, and then quantifies the electromagnetic interference level.

[0052] Through the third data channel, wind speed, vibration, and electromagnetic interference data are processed and fused in a unified manner. Data processing methods such as weighted average and principal component analysis are used to generate an environmental fusion feature set. This feature set integrates the influence of the three environmental factors and can reveal the comprehensive risk level under different environmental conditions (such as high wind speed, strong vibration, and strong electromagnetic interference).

[0053] Multidimensional correlation analysis is used to reveal which environmental factors have a significant impact on certain specific work behaviors. For example, strong winds pose a greater risk to hoisting operations, and high vibrations are more dangerous for high-altitude operations. Through multidimensional analysis, the most dangerous combinations of work behaviors are identified under various environmental conditions, and an environmental behavior correlation matrix is ​​constructed. The rows and columns of the matrix correspond to environmental factors and work behaviors, respectively. A certain position in the matrix represents "the degree of danger of the worker being in a standing position at a wind speed of Q". This environmental behavior correlation matrix can quantify the correlation between environmental factors and behaviors and provide data support for subsequent safety assessments and risk warnings.

[0054] Safety assessment quantifies the safety status of operations under different environments by correlating environmental factors (wind speed, vibration, electromagnetic interference) and behavioral parameters. Based on the correlation between wind speed influencing factors and operational behavior, the specific impact of wind speed on safety is quantified, forming a wind speed safety status quantification parameter. This parameter will display the risk level of operations under different wind speed conditions. Similarly, by correlating the vibration hazard coefficient with operational behavior, the impact of vibration on operational safety is quantified, generating a vibration safety status quantification parameter. Based on the correlation between electromagnetic interference levels and operational behavior, the impact of electromagnetic interference on operational safety is quantified, generating an electromagnetic safety status quantification parameter.

[0055] In actual safety risk management, different environmental factors have varying degrees of impact on safety. Therefore, risk weight allocation assigns an appropriate weight to each environmental factor based on its actual impact, helping to comprehensively assess overall safety risks. Specifically, the impact of wind speed, vibration, and electromagnetic interference on safety is determined through analysis of environmental fusion feature sets, historical data, and field experience. The weight calculation uses the analytic hierarchy process (AHP) to construct a 3x3 judgment matrix, where matrix elements represent the importance ratios between factors. By calculating the maximum eigenvalue and corresponding eigenvector of the judgment matrix, the weight vectors for each factor are obtained, representing the weight coefficients for wind speed, vibration, and electromagnetic interference. A consistency check is then performed to ensure the consistency ratio is less than 0.1. The consistency index is calculated by subtracting the matrix order from the maximum eigenvalue and then dividing by the matrix order minus one. For each environmental factor, different weight coefficients are assigned based on its contribution to safety. For example, wind speed has a larger weight because changes in wind speed pose a direct threat to work operations; while electromagnetic interference has a smaller weight unless the work in that environment relies on highly sensitive electronic equipment.

[0056] Using pre-configured intelligent AI models, such as machine learning models and deep learning models, the system analyzes the quantitative parameters of wind speed safety status and combines them with corresponding wind speed weighting coefficients to conduct real-time assessments of the current wind speed conditions on site. This intelligent AI model dynamically updates the wind speed risk level based on real-time collected environmental data, promptly responding to potential safety hazards that wind speed changes may bring to work operations and generating a dynamic wind speed risk level. The dynamic wind speed risk level is a numerical value or level label used to represent the operational safety risk under the current wind speed conditions. The higher the wind speed or the closer it is to the safety threshold for workers, the higher the risk level.

[0057] Similar to wind speed risk levels, the intelligent AI model analyzes on-site vibration data, such as vibration intensity and frequency, and combines this with set vibration weighting coefficients to assess the operational risk under the current vibration condition and output a dynamic vibration risk level.

[0058] Similar to wind speed and vibration, the electromagnetic interference level at the site is monitored in real time. Combined with the corresponding electromagnetic interference weighting coefficient, the potential risks of electromagnetic interference to work safety are analyzed. In this process, electromagnetic data at the site, such as electromagnetic wave intensity and frequency, are collected, and AI algorithms are used to determine whether they pose a threat to work safety. Based on the dynamic monitoring of electromagnetic interference, a dynamic electromagnetic risk level is generated.

[0059] Furthermore, based on the environmental fusion feature set and the operational behavior parameters, a multidimensional correlation analysis is performed to construct an environmental behavior correlation matrix. The method includes:

[0060] Risk analysis is performed based on the operational behavior parameters to determine the hazard level of actions, and a behavior axis is defined based on the hazard level. Risk analysis is also performed based on the environmental fusion feature set to determine the comprehensive environmental index, and an environmental axis is defined based on the comprehensive environmental index. Risk analysis is performed based on the equipment data at the tower operation site to determine the equipment integrity, and an equipment axis is defined based on the equipment integrity. A three-dimensional decision cube is constructed by using the behavior axis as the X-axis, the environmental axis as the Y-axis, and the equipment axis as the Z-axis. The environmental fusion feature set and the operational behavior parameters are correlated and mapped to the three-dimensional decision cube to construct the environmental behavior correlation matrix.

[0061] Risk analysis based on job behavior parameters assesses the potential hazard of each action. By combining the complexity, hazard, and working environment of the action, each action is scored or assigned a risk level. For example, low-risk actions indicate simple actions that will not lead to sudden accidents, such as standing steadily or low-intensity operations; medium-risk actions indicate more complex actions or actions involving certain risk factors, such as working at heights or high-load work for a short period; high-risk actions indicate dangerous actions or operations with a high probability of accidents, such as crossing high-voltage power lines or using hazardous chemicals. The behavior axis is defined based on the assessed hazard level of the action and describes the risk level of the job behavior. Low-risk actions correspond to the low-value area of ​​the behavior axis, and high-risk actions correspond to the high-value area.

[0062] Risk analysis based on an environmental feature set integrates multiple environmental features to derive a comprehensive environmental index. This comprehensive environmental index is a holistic measure reflecting the safety of the current working environment. For example, factors such as excessive wind speed, excessive vibration, and strong electromagnetic interference can cause the comprehensive environmental index to rise. Based on the calculated comprehensive environmental index, an environmental axis is defined to represent the safety of the current working environment. A lower environmental index indicates a safe working environment, and the environmental axis corresponds to a low-risk area; a higher environmental index indicates a higher working environment risk, and the environmental axis corresponds to a high-risk area.

[0063] Equipment data includes information on all critical equipment at the tower operation site, such as tools, machinery, support devices, and protective equipment. Risk analysis is conducted based on this data to determine if there are any malfunctions, wear, damage, or performance degradation. The equipment risk analysis also includes whether the equipment has been properly maintained and repaired, and whether it meets the current operational requirements. Equipment axes are defined based on equipment integrity, reflecting the safety of equipment at the operation site. High integrity equipment corresponds to the low-risk area of ​​the equipment axis, indicating that the equipment is in good condition; low integrity equipment corresponds to the high-risk area of ​​the equipment axis, indicating that the equipment is malfunctioning or does not meet safety requirements.

[0064] A three-dimensional decision cube is constructed, in which each point represents a specific operational scenario, describing the comprehensive risks of operational behavior, environmental conditions, and equipment status under that scenario. The three-dimensional decision cube can visualize these three different risk factors (behavior, environment, and equipment) in a unified space, providing a clear framework for subsequent decision analysis.

[0065] The values ​​of the environmental fusion feature set and operational behavior parameters are mapped onto a constructed 3D decision cube. The environmental fusion feature set affects the value on the Y-axis (environment axis), the operational behavior parameters affect the value on the X-axis (behavior axis), and the equipment data affects the value on the Z-axis (equipment axis). The specific mapping process involves converting the operational behavior parameters, environmental fusion feature set, and equipment data into cube coordinate points using corresponding normalization mapping functions. The original data is mapped to a range of zero to one. The mapping function uses a minimum-maximum normalization method, which involves subtracting the minimum value from the original value and then dividing by the difference between the maximum and minimum values. This means that the combined situation of the three factors determines the position of each point in the cube. For example, the state of the operational behavior, environment, and equipment at a certain moment will be located at a specific point in 3D space, forming an environmental-behavior correlation matrix. Each item in the matrix represents the combined risk level of behavior, environment, and equipment under different operational scenarios.

[0066] Furthermore, based on the dynamic risk level, safety alarms are generated at the tower operation site, and alarm information is pushed to a remote terminal for risk management. The method includes:

[0067] Risk is assessed based on the dynamic risk level to obtain a target risk response level; core risk fields are extracted according to the target risk response level, and the core risk fields are encapsulated to obtain a risk data packet; the alarm execution module is activated based on the risk data packet to determine priority and multiple priorities are identified; the risk data packet is pushed to a remote terminal for risk management according to the multiple priorities.

[0068] By analyzing the dynamic risk level of the current work site, the target risk response level during the work process is determined, which is the severity of the risk in each work scenario. If the dynamic risk level is high, it indicates that the risk of the work scenario is relatively large, so it is marked as a high-priority risk scenario and triggers a stronger warning. The target risk response level is the degree of response based on this assessment result, which is used to indicate the risk level that needs to be dealt with immediately.

[0069] Based on the target risk response level, core risk fields are extracted to describe and encapsulate each risk event, including: risk location, used to mark the specific segment number and three-dimensional coordinates of the tower to help determine the exact location of the event; risk type, which classifies risks into three categories based on the source of different factors: behavioral, environmental, and equipment-related; and environmental snapshot, containing peak wind speed and a thumbnail of the vibration spectrum 10 seconds before the triggering time. These extracted core risk fields are then encapsulated to generate a risk data package. This risk data package contains all necessary risk information and can be efficiently transmitted by the system to the backend processing system or a remote terminal.

[0070] After the generated risk data packet is sent to the alarm execution module, it is analyzed and prioritized based on factors such as the target risk response level and risk type. Multiple priorities are set. For example, the emergency level (highest priority) requires immediate handling and usually involves high-risk scenarios, such as when the life safety of workers is threatened or equipment is about to malfunction; the higher level (second highest priority) is more serious and may cause significant losses or affect the progress of the operation, but it may not immediately interrupt the operation; the general level (low priority) is relatively minor and can wait for further evaluation or observation, without immediate handling.

[0071] Based on the prioritization determined in the previous step, the generated risk data packets are pushed to remote terminals for risk management. For example, for urgent risks, the packets are pushed to relevant management personnel and on-site workers so that immediate measures can be taken; for lower-priority risks, the packets are pushed to safety monitoring personnel or the operations management team for subsequent observation and handling. Upon receiving the pushed risk data packets, the remote terminals respond according to their priority, including suspending operations, on-site intervention, issuing alarms, and providing safety notifications.

[0072] In summary, the intelligent AI-based tower safety operation monitoring method provided in this application has the following technical effects:

[0073] By collecting data from the tower operation site, a multidimensional dataset is obtained. Based on the type of acquisition equipment, the data is hierarchically managed, enabling precise acquisition of various operation-related information, avoiding information redundancy, and improving monitoring accuracy and efficiency. The data is layered into multiple data channels, each focusing on different types of information, making data analysis more targeted and efficient. The first data channel detects and locks onto targets at the tower operation site, capturing personnel and objects in real time during the operation, ensuring accurate identification of operational behaviors. The second data channel identifies personnel behavior and determines operational behavior parameters, identifying specific behaviors in different operational scenarios, providing a reliable basis for subsequent safety assessments. The third data channel integrates environmental data. By analyzing and combining operational behavior parameters for safety assessment, the system can comprehensively evaluate the safety situation at the work site. Utilizing AI technology for intelligent identification and dynamic monitoring, it can update the risk level of the work site in real time, ensuring timely response to any changes in the environment and behavior, thereby improving operational safety. Based on the assessment results, a dynamic risk level is generated, and safety alarms are issued for the work site based on this level. Alarm information is pushed to remote terminals for risk management, ensuring that risk information reaches the responsible parties quickly and enabling timely response. The remote terminal takes corresponding control measures based on the risk level, effectively preventing potential safety accidents. Through real-time monitoring and alarm mechanisms, the safety of tower operations is significantly improved, the probability of accidents is reduced, and the lives of workers are protected.

[0074] Example 2, based on the same inventive concept as the intelligent AI-based tower safety operation monitoring method in the aforementioned examples, such as... Figure 2 As shown in the embodiment of this application, a tower safety operation monitoring system based on intelligent AI recognition is provided. The system includes:

[0075] The data acquisition module 10 is used to collect data from the tower operation site, obtain a multidimensional dataset, perform data layering on the multidimensional dataset, and determine multiple data channels. These multiple data channels correspond to the data layers and include a first data channel, a second data channel, and a third data channel. The personnel behavior recognition module 20 is used to perform target detection through the first data channel, and to perform personnel behavior recognition based on the locked target and the second data channel, determining operational behavior parameters. The safety assessment module 30 is used to perform environmental fusion analysis through the third data channel, and to perform a safety assessment based on the environmental fusion feature set and the operational behavior parameters, generating quantitative parameters for safety status monitoring and identification using intelligent AI, and generating a dynamic risk level. The safety alarm module 40 is used to issue safety alarms to the tower operation site based on the dynamic risk level, generating alarm information and pushing it to a remote terminal for risk management.

[0076] Furthermore, the data acquisition module 10 is used to perform the following operation steps:

[0077] Based on the multidimensional dataset, the data is layered according to the type of acquisition device to obtain an image acquisition data layer, a behavior perception data layer, and an environmental monitoring data layer. The image acquisition data layer transmits video frame sequences for compression and allocation to construct the first data channel. The behavior perception data layer transmits personnel posture data for calculation to construct the second data channel. The environmental monitoring data layer transmits environmental physical parameters for filtering to construct the third data channel.

[0078] Furthermore, the personnel behavior recognition module 20 is used to perform the following operation steps:

[0079] Based on the first data channel, video frame analysis is performed on the tower operation site according to the video frame sequence to obtain a video data stream; multiple target detection information of the tower operation site is retrieved, and based on the video data stream, the multiple target detection information is used for identification to obtain a locked target, the locked target including target bounding box coordinates; spatial positioning is performed according to the target bounding box coordinates to determine personnel spatial location information, and the personnel spatial location information is synchronized to the second data channel for personnel behavior recognition to determine the operation behavior parameters.

[0080] Furthermore, the personnel behavior recognition module 20 is used to perform the following operation steps:

[0081] The second data channel is used to identify personnel at the tower operation site according to the operation sequence, and obtain personnel skeletal joint time sequence data; based on the personnel skeletal joint time sequence data, time sequence is extracted to determine multiple time window segments; according to the multiple time window segments, the personnel spatial location information is synchronized to the second data channel and associated with the personnel skeletal joint time sequence data to determine the operation behavior parameters.

[0082] Furthermore, the personnel behavior recognition module 20 is used to perform the following operation steps:

[0083] The center coordinates of the target bounding box are extracted. Using the center coordinates as the center, the temporal data of the personnel's skeletal joints are normalized according to the multiple time window segments to determine a three-dimensional coordinate sequence. Graph convolution feature analysis is performed on the three-dimensional coordinate sequence to extract spatial topological features. Temporal convolution feature analysis is performed on the three-dimensional coordinate sequence to extract action evolution features. The spatial position information of the personnel is associated and matched with the temporal data of the personnel's skeletal joints according to the spatial topological features to obtain a first operational behavior parameter. The spatial position information of the personnel is associated and matched with the temporal data of the personnel's skeletal joints according to the action evolution features to obtain a second operational behavior parameter.

[0084] Furthermore, the security assessment module 30 is used to perform the following operational steps:

[0085] The environmental physical parameters are analyzed to generate wind speed influencing factors, vibration hazard coefficients, and electromagnetic interference levels. These factors are then fused and analyzed through the third data channel to construct an environmental fusion feature set. A multi-dimensional correlation analysis is performed between the environmental fusion feature set and the operational behavior parameters to construct an environmental behavior correlation matrix. The environmental behavior correlation matrix is ​​traversed to perform a safety assessment based on the wind speed influencing factors, vibration hazard coefficients, and electromagnetic interference levels, generating quantitative parameters for wind speed safety status, vibration safety status, and electromagnetic safety status. Based on these parameters, safety risk weights are allocated to determine multiple weight coefficients. The wind speed safety status quantitative parameters are then dynamically monitored and identified using intelligent AI according to these weight coefficients to generate a dynamic wind speed risk level. Similarly, the vibration safety status quantitative parameters are dynamically monitored and identified using intelligent AI according to these weight coefficients to generate a dynamic vibration risk level. Finally, the electromagnetic safety status quantitative parameters are dynamically monitored and identified using intelligent AI according to these weight coefficients to generate a dynamic electromagnetic risk level.

[0086] Furthermore, the security assessment module 30 is used to perform the following operational steps:

[0087] Risk analysis is performed based on the operational behavior parameters to determine the hazard level of actions, and a behavior axis is defined based on the hazard level. Risk analysis is also performed based on the environmental fusion feature set to determine the comprehensive environmental index, and an environmental axis is defined based on the comprehensive environmental index. Risk analysis is performed based on the equipment data at the tower operation site to determine the equipment integrity, and an equipment axis is defined based on the equipment integrity. A three-dimensional decision cube is constructed by using the behavior axis as the X-axis, the environmental axis as the Y-axis, and the equipment axis as the Z-axis. The environmental fusion feature set and the operational behavior parameters are correlated and mapped to the three-dimensional decision cube to construct the environmental behavior correlation matrix.

[0088] Furthermore, the security alarm module 40 is used to perform the following operation steps:

[0089] Risk is assessed based on the dynamic risk level to obtain a target risk response level; core risk fields are extracted according to the target risk response level, and the core risk fields are encapsulated to obtain a risk data packet; the alarm execution module is activated based on the risk data packet to determine priority and multiple priorities are identified; the risk data packet is pushed to a remote terminal for risk management according to the multiple priorities.

[0090] Through the foregoing detailed description of the intelligent AI-based tower safety operation monitoring method, those skilled in the art can clearly understand the intelligent AI-based tower safety operation monitoring system in this embodiment. Since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and relevant parts can be found in the method section.

[0091] Example 3 provides a storage medium on which a computer program is stored, which, when executed by a processor, implements any step of Example 1.

[0092] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0093] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for monitoring safe operation of iron towers using intelligent AI recognition, characterized in that, include: Data is collected from the tower operation site to obtain a multidimensional dataset. The multidimensional dataset is then layered to determine multiple data channels. These multiple data channels correspond to the data layers and include a first data channel, a second data channel, and a third data channel. Target detection is performed through the first data channel, and personnel behavior is identified by combining the locked target with the second data channel to determine the operational behavior parameters. Environmental fusion analysis is performed through the third data channel. Based on the environmental fusion feature set and the operational behavior parameters, a safety assessment is conducted to generate quantitative parameters of the safety situation. Intelligent AI risk dynamic monitoring and identification are then performed to generate dynamic risk levels. Based on the dynamic risk level, safety alarms are generated at the tower operation site, and alarm information is pushed to a remote terminal for risk management. Data is collected from the tower operation site to obtain a multidimensional dataset. The multidimensional dataset is then layered to determine multiple data channels. These data channels correspond to data layers and include a first data channel, a second data channel, and a third data channel. The method includes: Based on the multidimensional dataset, the data is layered according to the type of acquisition device to obtain an image acquisition data layer, a behavior perception data layer, and an environmental monitoring data layer. The first data channel is constructed by compressing and allocating video frame sequences transmitted through the image acquisition data layer; The second data channel is constructed by processing the human posture data transmitted through the behavior perception data layer. The environmental physical parameters are filtered through the environmental monitoring data layer to construct the third data channel; The method includes: performing target detection through the first data channel, and identifying personnel behavior based on the locked target in conjunction with the second data channel to determine operational behavior parameters. Based on the first data channel, video frame analysis is performed on the tower operation site according to the video frame sequence to obtain a video data stream; Multiple target detection information from the tower operation site is retrieved, and the target is identified based on the video data stream according to the multiple target detection information to obtain the locked target, wherein the locked target includes the target bounding box coordinates; Spatial positioning is performed based on the target bounding box coordinates to determine the personnel spatial location information. The personnel spatial location information is then synchronized to the second data channel for personnel behavior recognition to determine the work behavior parameters. Spatial positioning is performed based on the target bounding box coordinates to determine the personnel spatial location information. This personnel spatial location information is then synchronized to the second data channel for personnel behavior recognition to determine the operational behavior parameters. The method includes: The second data channel is used to identify personnel at the tower operation site according to the operation sequence, and obtain the time sequence data of personnel skeletal joints. Based on the temporal data of the human skeletal joints, time sequence segmentation is performed to determine multiple time window segments; The spatial location information of the personnel is synchronized to the second data channel according to the multiple time window segments and associated with and matched with the temporal data of the personnel's skeletal joints to determine the operation behavior parameters.

2. The intelligent AI-based tower safety operation monitoring method according to claim 1, characterized in that, The method involves synchronizing the spatial location information of the personnel to the second data channel according to the multiple time window segments and then associating and matching it with the temporal data of the personnel's skeletal joints to determine the operational behavior parameters. Extract the center coordinate point of the target bounding box, and use the center coordinate point as the center to normalize the temporal data of the human skeletal joints according to the multiple time window segments to determine the three-dimensional coordinate sequence. Based on the three-dimensional coordinate sequence, graph convolution feature analysis is performed to extract spatial topological features; Temporal convolutional feature analysis is performed based on the three-dimensional coordinate sequence to extract action evolution features; The spatial location information of the personnel is associated and matched with the temporal data of the personnel's skeletal joints according to the spatial topological features to obtain the first operational behavior parameters; The spatial location information of the personnel is associated and matched with the temporal data of the personnel's skeletal joints according to the action evolution characteristics to obtain the second operational behavior parameters.

3. The intelligent AI-based tower safety operation monitoring method according to claim 1, characterized in that, Environmental fusion analysis is performed through the third data channel. A safety assessment is conducted based on the environmental fusion feature set and the operational behavior parameters. Quantitative safety status parameters are generated for intelligent AI-based dynamic risk monitoring and identification, resulting in a dynamic risk level. The method includes: The environmental physical parameters are analyzed to generate wind speed influence factor, vibration hazard coefficient, and electromagnetic interference level; The wind speed influencing factor, the vibration hazard coefficient, and the electromagnetic interference level are fused and analyzed through the third data channel to construct an environmental fusion feature set. A multidimensional correlation analysis is performed based on the environmental fusion feature set and the operational behavior parameters to construct an environmental behavior correlation matrix. The environmental behavior correlation matrix is ​​traversed to perform a safety assessment based on the wind speed influence factor, the vibration hazard coefficient, and the electromagnetic interference level, generating quantitative parameters of wind speed safety status, quantitative parameters of vibration safety status, and quantitative parameters of electromagnetic safety status. Based on the wind speed safety situation quantification parameters, the vibration safety situation quantification parameters, and the electromagnetic safety situation quantification parameters, safety risk weights are allocated, and multiple weight coefficients are determined. Intelligent AI risk dynamic monitoring and identification is performed on the wind speed safety status quantification parameters according to the multiple weighting coefficients to generate a dynamic wind speed risk level. Intelligent AI risk dynamic monitoring and identification are performed on the vibration safety status quantification parameters according to the multiple weighting coefficients to generate a vibration dynamic risk level. The electromagnetic security situation quantification parameters are dynamically monitored and identified using intelligent AI based on the multiple weighting coefficients to generate an electromagnetic dynamic risk level.

4. The intelligent AI-based tower safety operation monitoring method according to claim 3, characterized in that, Based on the environmental fusion feature set and the operational behavior parameters, a multidimensional correlation analysis is performed to construct an environmental behavior correlation matrix. The method includes: Risk analysis is performed based on the aforementioned work behavior parameters to determine the hazard level of the action, and a behavior axis is defined based on the hazard level of the action. Risk analysis is performed based on the environmental fusion feature set to determine a comprehensive environmental index, and an environmental axis is defined based on the comprehensive environmental index. Risk analysis is performed based on the equipment data at the tower operation site to determine the equipment integrity, and equipment axes are defined based on the equipment integrity. A three-dimensional decision cube is constructed by taking the behavior axis as the X-axis, the environment axis as the Y-axis, and the equipment axis as the Z-axis. The environmental fusion feature set and the operational behavior parameters are associated and mapped to the three-dimensional decision cube to construct the environmental behavior association matrix.

5. The intelligent AI-based tower safety operation monitoring method according to claim 1, characterized in that, Based on the dynamic risk level, safety alarms are generated at the tower operation site, and alarm information is pushed to a remote terminal for risk management. The method includes: Based on the dynamic risk level, a risk assessment is performed to obtain the target risk response level; Extract the core risk fields according to the target risk response level, encapsulate the core risk fields, and obtain a risk data package; Based on the risk data packet, the alarm execution module is activated to determine the priority of multiple priorities. The risk data packets are pushed to remote terminals for risk management according to the multiple priorities.

6. A tower safety operation monitoring system based on intelligent AI recognition, characterized in that: The system is used to implement the intelligent AI recognition method for monitoring safe operation of iron towers according to any one of claims 1 to 5, the system comprising: The data acquisition module is used to collect data from the tower operation site, obtain a multidimensional dataset, perform data layering on the multidimensional dataset, determine multiple data channels, the multiple data channels correspond to the data layers, and the multiple data channels include a first data channel, a second data channel, and a third data channel. The personnel behavior recognition module is used to perform target detection through the first data channel, and to perform personnel behavior recognition based on the locked target and the second data channel to determine the operation behavior parameters. The safety assessment module is used to perform environmental fusion analysis through the third data channel, conduct safety assessment based on the environmental fusion feature set and the operation behavior parameters, generate safety situation quantitative parameters for intelligent AI risk dynamic monitoring and identification, and generate dynamic risk levels. The safety alarm module is used to issue safety alarms at the tower operation site based on the dynamic risk level, and generate alarm information to push to remote terminals for risk management.

7. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the intelligent AI recognition method for monitoring safe operation of iron towers as described in any one of claims 1 to 5.

Citation Information

Patent Citations

  • Construction safety real-time high-precision detection method and system based on environmental characteristics

    CN120375294A