Concrete construction process and worker health monitoring system based on computer vision

Abstract

The invention discloses a concrete construction process and worker health monitoring system based on computer vision. The system comprises an input module, a calculation module and an output module. The input module collects construction site pictures in real time and carries out real-time processing through embedded equipment. The calculation module receives data collected by the input module, wherein the data comprise the concrete worker posture, the equipment operation state and the specific construction task requirement. The system utilizes improved YOLOv8-D and YOLOv8-P algorithms to accurately detect attitude characteristics of construction equipment and concrete workers. Besides, the system classifies and identifies construction operation categories by supporting an SVM, analyzes continuous multi-frame image data in combination with ST-GCN, and captures dynamic changes of worker postures and equipment states in a construction site. And the output module has the functions of construction process analysis and health monitoring. The system feeds back the construction state and early warning information to an operator through a human-computer interaction interface in a voice broadcast mode, helps the operator to adjust the operation mode in time, and ensures the safety and high efficiency of the construction process.

Description

Technical Field

[0001] The present invention belongs to the field of building information technology and intelligent construction equipment, and specifically relates to a concrete construction process and worker health monitoring system based on computer vision. Background Art

[0002] With the rapid advancement of infrastructure construction, concrete, as the most commonly used material in construction projects, has a construction quality and efficiency that is directly related to the overall progress and safety of the project. However, traditional concrete construction relies on manual operation and monitoring, which is not only inefficient, but also prone to quality problems due to errors in pouring and vibrating. At the same time, construction workers maintaining a single posture or repeating movements for a long time can easily cause health risks, further affecting construction progress and personnel efficiency. Therefore, realizing the automation and intelligence of the construction process is of great significance to improving construction quality and health monitoring levels.

[0003] In recent years, computer vision and deep learning technologies have developed rapidly. For example, the application of target detection, posture estimation, and time series analysis algorithms has provided a new means for intelligent monitoring of construction sites. These technologies make it possible to capture the movements of concrete workers in real time, monitor the operating status of equipment, and analyze dynamic construction scenes. Although the existing system can perform basic identification of concrete workers and equipment, it still has shortcomings in the real-time nature of process monitoring and the accuracy of health monitoring. The automated monitoring system that combines the YOLOv8, SVM, and ST-GCN algorithms can achieve a comprehensive analysis of construction scenes and dynamic monitoring of personnel health status, helping to comprehensively improve construction process management and safety assurance. Summary of the invention

[0004] To solve the above problems, the present invention discloses a concrete construction process and worker health monitoring system based on computer vision, which integrates image processing technology, posture detection and timing analysis technology. By real-time analysis of image data of the construction site and the posture behavior of workers, the system can automatically identify and optimize the operation process of concrete construction, thereby improving construction quality and workers' health level.

[0005] To achieve the above object, the technical solution of the present invention is as follows:

[0006] The computer vision-based concrete construction process and worker health monitoring system includes an input module, a calculation module, and an output module.

[0007] Furthermore, the input module is responsible for receiving and processing data information of the construction site, including personnel postures, equipment operating status and their interactive behaviors, and realizing intelligent identification and monitoring of the construction site through real-time data collection and transmission.

[0008] Furthermore, the specific functions of the input module are:

[0009] As the core of data collection, the input module is responsible for the real-time acquisition and transmission of construction site information, providing basic data support for subsequent analysis. The module collects construction images through embedded devices and high-definition cameras, comprehensively records the status information of concrete workers and equipment, and ensures real-time grasp of the dynamics of the site. Specifically, the input module can capture the body movements and posture information of concrete workers, such as key actions such as standing, bending, and raising hands. In addition, the input module can also identify the interactive behaviors between people and equipment, such as the dynamic process of operating a concrete placing machine or a mixer, and synchronously collect multi-channel data, such as video images and environmental parameters, to ensure the consistency of data in the time dimension. Through the above functions, the input module provides a high-quality data foundation for process monitoring and health management of the construction site.

[0010] The input module supports multiple input methods, can monitor the construction site through image processing technology, and process the posture and equipment operation data collected on site in combination with artificial intelligence technology.

[0011] Furthermore, the computing module receives the data information collected by the input module, classifies and identifies the construction operations based on the Support Vector Machine (SVM) classification model and the Spatial-Temporal Graph Convolutional Network (ST-GCN), monitors the construction process and determines whether an early warning is required, as follows:

[0012] 1. Construction image processing. The system uses the embedded device JetsonNano camera to collect construction site images in real time, and uses YOLOv8 image processing technology to detect and locate key targets. The detection process includes three stages: feature extraction, feature fusion, and target prediction:

[0013] (1) In the feature extraction stage, the input image I is extracted through the backbone network to extract preliminary features F:

[0014] F=Backbone(I)

[0015] The backbone network extracts the spatial and semantic information of the image through multi-layer convolution operations to provide feature support for subsequent detection;

[0016] (2) In the feature fusion stage, the system fuses features of different scales through the feature fusion module to generate a multi-scale feature map F fusion

[0017] F fusion =Neck(F)

[0018] This step strengthens the multi-scale detection capabilities of workers, concrete placing machines and mixers by combining multi-level information, ensuring detection accuracy in complex construction environments;

[0019] (3) In the target prediction stage, the prediction module is used to output the target’s category probability P, bounding box parameters B = (x, y, w, h) and confidence score C; the detection result D is expressed as:

[0020] D={(P i ,B i ,C i )|i=1,2,...,N}

[0021] Where N is the number of detected targets, P i is the target category, B i Represents the bounding box of the target, including the center coordinates (x, y) and size (w, h), C i represents the confidence score;

[0022] The system further uses the non-maximum suppression algorithm to remove redundant detection boxes and only retains the target box B with the highest confidence. max :

[0023]

[0024] Among them, IoU(B i ,B j ) is the bounding box B i and B j The intersection-and-union ratio is used to quantify the degree of overlap of the target area. Through the above process, the system can accurately detect and locate workers, placing machines and mixers, provide real-time monitoring and support for key activities in the construction process, and effectively improve construction efficiency and safety.

[0025] 2. Construction activity monitoring. The system uses the SVM classification model to identify the detected action categories, such as concrete pouring and vibration operations. SVM is based on the input posture feature vector X, through the hyperplane function f(X) = ω T X+b determines the action category and combines it with ST-GCN to perform dynamic behavior analysis. The system further compares the analysis results with the time requirements of the construction process. For example, the duration of the concrete pouring operation should not exceed 90 minutes, and the single vibration operation time should be kept within the range of 5 to 15 seconds. When it is detected that the actual duration of a process exceeds the specified standard or the operation frequency deviates from the standard value, the system will generate an abnormal warning of the construction activity, reminding the construction personnel to adjust the operation or optimize the construction rhythm to ensure that the construction task proceeds as planned.

[0026] 3. Health monitoring. The system calculates the elbow angle and movement frequency of concrete workers in real time based on posture estimation technology. By calculating the angle θ between the elbow and the shoulder, the formula Determine whether the action exceeds the healthy threshold. Points A, B, and C represent the positions of the shoulder, elbow, and wrist, respectively. The vector and They are the directions from shoulder to elbow and shoulder to wrist respectively; ST-GCN captures the interaction between concrete workers and equipment through graph convolution operations, maps multi-frame data into time series features, and further evaluates whether the duration and frequency of the operation meet the specifications. When the elbow angle θ>150° or the hand-raising frequency f>20 times / minute, the system generates a health warning signal, and reminds the concrete worker to adjust the operation through voice broadcast and interface prompts to avoid health problems caused by repetitive actions.

[0027] 4. Construction activity monitoring and feedback. The system dynamically tracks the actual progress of construction tasks. When the operation time exceeds the preset threshold or unhealthy operations are detected, the system issues an early warning and displays specific abnormal information through the human-computer interaction interface to ensure the improvement of construction quality and efficiency.

[0028] Furthermore, the function of the output module is to provide real-time feedback to construction operators through a human-computer interaction interface and a voice broadcast system to ensure that they adjust their operations in a timely manner to ensure the safety and efficiency of construction, while reducing health risks caused by improper operation.

[0029] The method for using the computer vision-based concrete construction process and worker health monitoring system of the present invention comprises the following steps:

[0030] 1. On-site scene analysis: The input module processes the concrete construction site video obtained by the embedded device through computer vision technology, including gesture recognition and image analysis, and extracts the core information of concrete construction, including but not limited to personnel location, equipment status and key indicators of construction activities;

[0031] 2. Construction status identification and analysis: The system uses YOLOv8 and SVM models, combined with ST-GCN for time series analysis, to monitor the actions of construction site personnel and equipment status in real time. It not only generates detailed construction status information, but also detects abnormal behaviors, ensuring that every link of the construction process is accurately monitored;

[0032] 3. Early warning feedback: The system determines whether an early warning is needed for the construction process or the health of workers based on the calculation results. The early warning information provides feedback to the operator through intuitive voice broadcast and interface display to ensure the smooth progress of the construction process and protect the health of workers.

[0033] The beneficial effects of the present invention are:

[0034] 1. Immediacy and convenience. Through embedded devices, managers can obtain the concrete construction status through this system anytime and anywhere, which improves the flexibility and response speed of management.

[0035] 2. Accuracy. By combining the SVM model, YOLOv8 and ST-GCN, the system can monitor the construction site in real time, ensuring the automation and intelligence of the construction process and health monitoring; at the same time, the system's high-precision detection capability can accurately capture the dynamic changes in human-computer interaction, and maintain high efficiency and high reliability even in complex and changing construction environments, thereby significantly improving the accuracy of abnormal posture recognition; effectively avoiding misunderstandings or information distortion caused by human factors.

[0036] 3. Health protection. The system can monitor the health of workers in real time, especially the detection of repetitive actions such as frequent hand raising, which helps prevent occupational diseases and protect the health of workers.

[0037] 4. Improve efficiency. The real-time feedback and guidance provided by the system help operators adjust operations in a timely manner, optimize construction processes, and improve work efficiency. This improvement in efficiency not only speeds up the construction progress, but also reduces the costs caused by delays and rework, saving time and resources for the project.

[0038] 5. 24 / 7 service. The invention provides all-weather service. This uninterrupted service ensures the continuity and timeliness of construction management. Even at night or in bad weather conditions, the system can work normally and provide managers with the required information. This all-weather monitoring and feedback mechanism provides additional security for construction projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a framework diagram for implementing the present invention.

[0040] Figure 2 Worker posture diagram. DETAILED DESCRIPTION

[0041] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention.

[0042] As shown in the figure, the computer vision-based concrete construction progress and worker health monitoring system of the present invention includes:

[0043] 1. Input module

[0044] The input module is designed to adapt to the complex environment of the construction site and integrates embedded devices such as cameras and Jetson Nano. The system captures the construction site in real time through a high-definition camera and uses the YOLOv8 algorithm for worker posture recognition and equipment detection. The main function of this module is to extract key construction information from the image, such as workers, equipment, and their spatial coordinates.

[0045] 2. Computing module

[0046] The computing module uses SVM combined with YOLOv8-P and YOLOv8-D algorithms to classify and identify construction tasks in real time, monitor the duration of concrete pouring and vibration operations, and perform time series data analysis through ST-GCN. The specific functions are as follows:

[0047] a. Equipment detection. The computing module uses the YOLOv8 model to detect equipment and identify construction scenes on images of the construction site, providing key data support for subsequent construction activity analysis and operation plan formulation. The system uses YOLOv8 to detect targets such as concrete workers, concrete placing machines, and vibrators, and extracts their category labels and location coordinates, including center point coordinates (x, y) and size parameters (w, h)

[0048] First, in the model training phase, the system uses the binary cross entropy loss function to optimize the YOLOv8 model, and uses the Adam optimizer to dynamically adjust the learning rate and momentum parameters to ensure that the model can converge efficiently and improve detection accuracy. Subsequently, in the inference phase, the input construction site image is extracted through the YOLOv8 backbone network, and then multi-scale feature integration is performed through the feature fusion module, and finally the category and location information of the target is generated in the prediction module.

[0049] Then, in the post-processing stage, the system uses the non-maximum suppression method to remove redundant detection frames and retain only the frames with the highest confidence, thereby improving the reliability and accuracy of the detection results. Finally, the system outputs the category labels of workers and equipment and their corresponding spatial location information as structured data to support subsequent construction task analysis and solution optimization.

[0050] b. Construction activity monitoring. Through multi-level analysis of the dynamic data of the construction site, process identification and status monitoring are realized. The system first detects the key targets in the construction task through the YOLOv8 model, including the status and position of concrete workers and equipment such as concrete placing machines and vibrators, and extracts their category labels and location information. Subsequently, the system combines SVM to classify the detected process categories and clarify different operation types such as concrete pouring and vibration. SVM makes judgments based on the posture feature vector and uses the decision function f(X)=w T X+b determines the operation category.

[0051] In order to further capture the dynamic characteristics of the process, the system introduces ST-GCN to perform time series analysis on continuous frame data. ST-GCN deeply analyzes key indicators such as the duration and frequency of the process by modeling the interaction between concrete workers and equipment. For example, the system can monitor whether the single duration of the concrete pouring operation exceeds 90 minutes, and whether the single duration of the vibration operation remains within the prescribed range of 5 to 15 seconds. When it is detected that the continuous use of the equipment exceeds the safety limit or the worker's operation has non-compliant behavior, the system will generate an abnormal warning. At the same time, the structured process information generated by the system can not only evaluate the status of the construction task, but also provide a basis for optimizing the construction plan and further improve construction efficiency.

[0052] c. Health warning. By analyzing the posture information and movement characteristics of concrete workers in real time, the system can realize dynamic monitoring and warning functions of health status. The system first calculates key indicators such as elbow angle and movement frequency based on the posture data of concrete workers. By extracting the key point positions of the shoulder, elbow and wrist, the formula for calculating the elbow angle θ is: in, and Represent the vectors from shoulder to elbow and shoulder to wrist respectively. When the elbow angle θ>150° or the hand raising frequency f>20 times / minute, the system determines it as a high-risk action and generates a health warning signal.

[0053] The system combines ST-GCN to perform time-series analysis on the dynamic behavior of concrete workers, evaluating the duration of actions and the frequency of repetitive operations. When it detects that the posture or frequency exceeds the set threshold, such as maintaining a single action for a long time or raising the hand too frequently, the system reminds the concrete worker to adjust the operation in time through voice broadcast and interface prompts to avoid muscle strain or fatigue caused by excessive operation.

[0054] 3. Output module

[0055] The output module is responsible for feeding back the construction activity inspection results and worker health monitoring results generated by the calculation module to the administrator in the form of text and voice. The specific functions are as follows:

[0056] a. Text display. The system will display the construction process and worker health check results in text form on the user interface, and the administrator can directly view abnormal construction activities and warning information.

[0057] b. Voice broadcast: To help concrete construction workers obtain information when they cannot view the screen during operation, the system supports voice broadcast through speakers, providing real-time feedback on the construction process and health warning prompts.

[0058] System physical architecture:

[0059] 1. User interface layer

[0060] User interface: Provides an intuitive, easy-to-use interface for managers to view construction progress and warning information.

[0061] Input devices: including embedded devices such as Jetson Nano cameras, used to receive images from the construction site.

[0062] Output devices: such as display screens, speakers, etc., used to display process progress or provide voice broadcasts of health reminders.

[0063] 2. Infrastructure layer

[0064] Embedded device Jetson Nano camera: used to collect construction site images and perform preliminary processing.

[0065] Servers and storage devices: provide computing resources and data storage space to ensure stable operation of the system.

[0066] Network equipment: ensures smooth data communication between various components and supports real-time data transmission and feedback.

[0067] It should be noted that the above content only illustrates the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. For ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications all fall within the protection scope of the claims of the present invention.

Claims

1. Computer vision-based concrete construction process and worker health monitoring system, characterized by: Including input module, calculation module, output module, Its working principle is: based on computer vision technology, it realizes comprehensive monitoring of the concrete construction process and accurate monitoring of the health status of workers, collects construction site data in real time through embedded devices Jetson Nano and cameras, and uses the YOLOv8 algorithm to detect key targets, including workers, concrete placing machines and mixers, and ensures the accuracy and robustness of the detection results through feature fusion and non-maximum suppression technology; the system further integrates the SVM classification model and the ST-GCN timing analysis model to realize the classification and dynamic behavior monitoring of construction actions, and evaluate the compliance of construction duration and frequency; in terms of health monitoring, the system combines limb key point detection and dynamic analysis to provide real-time warnings; through the human-computer interaction interface, the system transmits construction and health information to the operator in the form of voice broadcast and visual feedback, effectively improving construction efficiency, safety and worker health management level.

2. The computer vision-based concrete construction process and worker health monitoring system according to claim 1 is characterized by: The input module is responsible for receiving data information from the construction site, including image data collected by the camera and related environmental parameter data, and transmitting these data to the calculation module for further processing and analysis.

3. The computer vision-based concrete construction process and worker health monitoring system according to claim 2 is characterized in that: The specific function of the input module is that the system uses the YOLOv8 algorithm to identify key entities and location information in the construction task, and converts this information into structured query data for further analysis by the calculation module.

4. The computer vision-based concrete construction process and worker health monitoring system according to claim 2 is characterized in that: The input module collects construction site images and videos through cameras, and processes the posture and equipment operation data collected on site in combination with artificial intelligence technology.

5. The computer vision-based concrete construction process and worker health monitoring system according to claim 1, characterized in that: The calculation module receives the data information collected by the input module, and classifies and identifies the construction operations based on the SVM classification model and ST-GCN, monitors the construction process and determines whether an early warning is needed, as follows: a. Understanding of the construction scene This system is based on the embedded device Jetson Nano. It uses a camera to collect construction site images in real time and uses YOLOv8 image processing technology to detect and locate key targets. The entire detection process is divided into three stages: feature extraction, feature fusion, and target prediction. In the feature extraction stage, the backbone network extracts the spatial and semantic information of the image through convolution kernel combinations, activation functions, and pooling layers, providing basic support for detection; In the feature fusion stage, the feature pyramid network is used to enhance the multi-scale detection capabilities of workers, material placing machines and mixers, ensuring detection accuracy in complex environments. In the target prediction stage, the prediction module outputs the target category, bounding box and confidence, and removes redundant boxes through the non-maximum suppression algorithm, retaining only the results with the highest confidence. The system can accurately detect and locate key targets on the construction site, provide real-time monitoring support for construction activities, and effectively improve construction efficiency and safety. b. Construction process monitoring The system uses the SVM classification model to identify the detected action categories, such as concrete pouring and vibration operations; SVM determines the action category based on the input posture feature vector X, and combines ST-GCN to perform dynamic time series analysis on the construction behavior; the system extracts the duration and frequency of the action, and compares the analysis results with the time requirements of the construction process; among them, the duration T of the concrete pouring operation should meet the range of T1≤90min, and the duration T2 of a single vibration operation should meet the range of 5s≤T2≤15s; otherwise, the system will trigger an abnormal warning of the construction activity, reminding the construction personnel to adjust the operation strategy; the warning information is displayed through voice prompts and interface visualization, helping construction personnel to optimize the construction rhythm and ensure that the task is completed as planned, while improving the construction safety and quality management level; c. Health monitoring The system can monitor the posture information of concrete workers in real time, including elbow extension and frequent hand raising operations, especially focusing on these high-risk actions to prevent occupational health problems caused by long-term repetitive operations. Based on the YOLOv8-Pose algorithm, the system first extracts the coordinates of the key points of the concrete workers, including the positions of the shoulders, elbows, and wrists. By calculating the angle θ of the elbow, the system determines whether the elbow is in an extended or hand-raising state. Specifically, the calculation formula of the angle θ is: Points A, B, and C represent the positions of the shoulder, elbow, and wrist, respectively. The vector and They are the directions from shoulder to elbow and from shoulder to wrist respectively; when the elbow angle θ>150°, the system determines that the elbow is in an overstretched state; In addition, the YOLOv8-Pose algorithm will record every time the worker raises or lowers his hand, thereby counting the number of times the concrete worker raises his hand per minute f; through the ST-GCN algorithm, if the action frequency f is detected to exceed the preset health threshold of 20 times / minute, the system will trigger a health warning to remind the concrete worker to take a rest and prevent fatigue or muscle strain due to excessive operation.

6. The computer vision-based concrete construction process and worker health monitoring system according to claim 1, characterized in that: The output module realizes multi-level construction status feedback function through the human-computer interaction interface, including health warning, operation guidance and risk warning. First, for the health risk monitoring of concrete workers, the system dynamically evaluates the operation status by detecting the elbow angle and movement frequency. When it is detected that the elbow angle exceeds the set range or the movement frequency exceeds the health threshold, the system immediately generates an early warning and transmits the early warning information to the operator through voice broadcast and visual interface, prompting the operator to take a rest or adjust the operation mode to avoid occupational health problems caused by long-term repetitive operation. The system's health monitoring and feedback mechanism is data-driven. It effectively improves the intelligence level of health management at the construction site through precise monitoring of elbow extension and frequent hand-raising behaviors, combined with real-time voice and visual feedback. At the same time, the mechanism uses dual warning methods of voice broadcast and interface prompts to ensure that operators can respond quickly and adjust irregular postures or movements while ensuring accurate communication of information. Ultimately, the system not only achieves health risk prevention and control during the construction process, but also provides reliable support for improving the overall operational safety of concrete workers.

7. The method for using the computer vision-based concrete construction process and worker health monitoring system according to claim 1 is characterized in that The following steps are involved: a. On-site scene analysis: Video data of the construction site is collected through embedded devices, and the input module uses computer vision technology to pre-process the image, including target detection and key point extraction; the system uses the YOLOv8 algorithm to identify concrete workers, concrete placing machines and mixers at the construction site, and records their position coordinates and status; at the same time, combined with posture estimation technology, the body movement information of the concrete workers is captured to provide multi-dimensional data support for subsequent analysis; b. Construction status recognition and analysis: The system uses YOLOv8 target detection, SVM classification model and ST-GCN time series analysis model to recognize and analyze the dynamic behavior of the construction site. First, the detected concrete pouring and vibration action categories are classified through the SVM model. Then, ST-GCN monitors the interaction between personnel and equipment and the changing trend of continuous actions based on the time series characteristics of multi-frame images. The system converts the analysis results into construction status information, including construction procedures, personnel operation status and equipment operation status, to provide support for real-time monitoring. c. Early warning feedback: Based on the analysis results, the system evaluates the health status and operational compliance of concrete workers; When abnormal postures are detected, including elbow angles exceeding 150°, hand-raising frequency exceeding 20 times per minute, or a single concrete pouring operation lasting more than 90 minutes, and a single vibration operation time not maintained within the prescribed range of 5 to 15 seconds, the system will generate an early warning signal; the early warning information will be conveyed to the concrete workers in real time through voice broadcasts, and specific problems will be displayed visually on the interface, including the type, frequency and time limit of abnormal movements; operators can adjust their operating behaviors, optimize the operating rhythm or take appropriate rest in a timely manner based on feedback information to ensure that construction tasks are carried out as required while reducing occupational health risks.

Images