A monitoring method and system for dynamic three-dimensional risk areas in bridge construction based on a monocular 3D detection algorithm

Abstract

The present invention discloses a monitoring method and system for dynamic three-dimensional risk areas in bridge construction based on a monocular 3D detection algorithm. The method includes collecting video images of the construction area through a monocular camera; using a trained 3D detection model to identify the three-dimensional spatial positions and sizes of construction machinery and operators; performing optical flow detection on the video images to determine the motion states of the construction machinery; generating and adjusting a dynamic three-dimensional risk area around the construction machinery that can move with the construction machinery based on the detection results; judging whether a person is an authorized operator by comparing face recognition with an identity database; monitoring the positions of the operators. When the construction machinery is running, an alarm is triggered when a person other than the operating driver enters the three-dimensional risk area. When the construction machinery is stationary, an alarm is triggered when an unauthorized team operator enters the three-dimensional risk area. Advantages: It can provide reliable technical support for safety management at the construction site, significantly improving the real-time performance, accuracy and adaptability of construction safety monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of highway construction risk monitoring, and particularly to a method and system for monitoring dynamic three-dimensional risk areas of bridge construction based on a monocular 3D detection algorithm. Background Art

[0002] Bridge construction, especially large-scale engineering projects such as cross-sea bridges, faces problems such as great construction difficulty, complex environment, and narrow site. To build the main bridge, it is usually necessary to erect a temporary sea trestle as a passage and construction platform, making the construction platform space narrow. There are a large number of construction machinery and operating personnel in the temporary sea trestle and the construction site area on the bridge deck, and they move frequently, which is prone to spatial intersection, resulting in a significant increase in safety risks.

[0003] In the safety monitoring of bridge construction areas, traditional monitoring methods mainly rely on two-dimensional image detection technology or manual monitoring by personnel, specifically including video monitoring, sensor detection, GPS positioning, etc. These methods have the following deficiencies:

[0004] 1. Limited ability to judge spatial positions:

[0005] Since two-dimensional image detection can only obtain planar information and cannot identify the precise positional relationship between construction machinery and operating personnel in three-dimensional space, it is difficult to accurately judge the relative distance between machinery and personnel and effectively monitor potential spatial intersection risks.

[0006] 2. Inaccurate division of dynamic risk areas:

[0007] Traditional monitoring methods usually rely on manual setting of safety distances and cannot dynamically adjust the risk area according to the real-time position changes of construction machinery and personnel and the changes in the motion state of construction machinery. In the case of frequent movement of construction machinery, it is easy to have a problem of mismatched risk area range, increasing potential safety hazards.

[0008] 3. Lack of real-time performance and adaptability:

[0009] In a complex environment such as a bridge construction site, the movement of machinery and personnel is frequent and the paths are not fixed. Traditional monitoring technologies are difficult to track and identify the specific positions of all machinery and personnel in real time and flexibly, thus making it difficult to provide accurate risk warning information.

[0010] 4. High hardware cost:

[0011] Some existing technologies (such as lidar) can provide three-dimensional space information, but their equipment costs are high and the installation is complex, and they are not suitable for large-scale application in high-density and dynamic construction scenarios. Summary of the Invention

[0012] The purpose of the present invention is to provide a method and system for monitoring dynamic three-dimensional risk areas during bridge construction based on a monocular 3D detection algorithm, so as to solve the foregoing problems existing in the prior art.

[0013] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0014] A method for monitoring dynamic three-dimensional risk areas during bridge construction based on a monocular 3D detection algorithm includes the following steps:

[0015] S1. Image acquisition: Real-time acquisition of video images of the construction area through a monocular camera;

[0016] S2. Image processing and 3D detection: Using a trained 3D detection model to identify the three-dimensional spatial positions and sizes of construction machinery and operating personnel at the construction site;

[0017] S3. Optical flow detection: Performing optical flow detection on the video images to analyze the motion state of the construction machinery;

[0018] S4. Risk area generation: Dynamically generating and adjusting a three-dimensional risk area surrounding the construction machinery that moves with the construction machinery based on the detection results of the 3D detection model and the optical flow detection results;

[0019] S5. Identity recognition: Real-time identification of the identity of personnel entering the risk area through face recognition and comparison with the identity database to confirm whether they are authorized operating personnel of the operation team of the construction machinery;

[0020] S6. Early warning: Real-time monitoring of the positions of operating personnel. When the construction machinery is running, an alarm will be triggered if a person other than the operating driver enters the three-dimensional risk area; when the construction machinery is stationary, an alarm will be triggered if a person other than the authorized team operating personnel enters the three-dimensional risk area.

[0021] Preferably, step S1 specifically includes the following content:

[0022] S11. Select a monocular camera with high resolution and wide viewing angle to ensure coverage of the main construction area;

[0023] S12. Install it at a high position to avoid occlusion and ensure that the camera can monitor key construction positions within the maximum range;

[0024] S13. The monocular camera real-time acquires video images of the construction area.

[0025] Preferably, the 3D detection model is built using the MMDetect ion3D framework; the training process of the 3D model is as follows:

[0026] S21. Acquire image data of various construction machinery and operating personnel, and mark the three-dimensional positions and sizes to construct a training data set for the construction scene;

[0027] S22. Use ResNet101 as the backbone. First, perform multi-level feature extraction on the image data, and then pass these features to the 3D detection head to complete specific tasks. Introduce DCN to enable the backbone to more flexibly adapt to the geometric deformation of the objects in the image, thereby enhancing the effect of feature extraction.

[0028] S23. Combine with the 3D detection head to predict the three-dimensional information and object category of the object. The three-dimensional information includes the value of the center offset in the x direction, the value of the center offset in the y direction, depth information, the width of the object, the length of the object, the height of the object, and the rotation angle of the object.

[0029] S24. Adopt supervised learning technology to train the 3D detection model so that it can predict the three-dimensional spatial positions and sizes of construction machinery and operators from the video images obtained by the monocular camera.

[0030] Preferably, step S3 specifically includes the following contents.

[0031] S31. Input two consecutive frames of images into the LiteFlowNet2 network. The LiteFlowNet2 network estimates the optical flow information layer by layer through a cascaded convolutional neural network, estimates the pixel displacements in the entire image from coarse to fine, and outputs the optical flow field.

[0032] S32. Use the 2D object detection model Faster R-CNN to perform construction machinery object detection, output the 2D object detection bounding box, and extract the optical flow data of the target area from the optical flow field according to the 2D object detection bounding box.

[0033] S33. Judge whether the motion state of the construction machinery is "moving" or "stationary" by calculating the average magnitude of the optical flow vectors within the target object area.

[0034] Preferably, step S4 specifically includes the following contents.

[0035] S41. Set the risk area expansion factor, which is used to determine the size of the risk area.

[0036] S42. Set the value of the risk area expansion factor according to the motion state of the construction machinery.

[0037] S43. Extract the basic tensor data of the construction machinery bounding box, including position coordinates, length, and width.

[0038] S44. Use the risk area expansion factor to expand the specified tensor data.

[0039] Preferably, step S42 is specifically as follows. When the construction machinery is in the "moving" state, the risk area expansion factor is set to 1, so that the size of the generated risk area is equivalent to 1 times the distance of the construction machinery bounding box; when the construction machinery is in the "stationary" state, the risk area expansion factor is set to 0.5, so that the size of the generated risk area is reduced to 0.5 times the distance of the construction machinery bounding box.

[0040] Preferably, step S44 is specifically as follows. Only the length and width parameters of the construction machinery bounding box are proportionally adjusted, and a new three-dimensional risk area bounding box is generated using the adjusted tensor data;

[0041] Multiply the length and width in the tensor data by (1 + expansion factor) respectively to achieve the expansion operation, while the height remains unchanged.

[0042] Preferably, step S5 specifically includes the following contents.

[0043] S51. Collect and register the facial features of the operating team workers, generate feature codes and store them in the identity database;

[0044] S52. Real-time capture the images of the personnel entering the risk area;

[0045] S53. Use the face detection model and face recognition model to identify and extract the face feature vectors of the personnel images, compare the face feature vectors of the personnel with all the feature vectors in the identity database, calculate the similarity between the two, and determine whether the corresponding personnel are authorized personnel.

[0046] The purpose of the present invention also lies in providing a bridge construction dynamic three-dimensional risk area monitoring system based on a monocular 3D detection algorithm, including:

[0047] Image acquisition module: Real-time collect the video images of the construction area through a monocular camera;

[0048] Image processing and 3D detection module: Use the trained 3D detection model to identify the three-dimensional spatial positions and sizes of the construction machinery and operating personnel at the construction site;

[0049] Optical flow detection module: Perform optical flow detection on the video images to analyze the motion state of the construction machinery;

[0050] Risk area generation: Dynamically generate a three-dimensional risk area around the construction machinery based on the detection results of the 3D detection model and the optical flow detection results;

[0051] Identity recognition module: Through face recognition and comparison with the identity database, real-time identify the identity of the personnel entering the risk area and confirm whether they are authorized operating personnel of the construction machinery operation team;

[0052] Early warning module: It monitors the positions of operators in real time. When the construction machinery is running, an alarm will be triggered if a person other than the operator enters the three-dimensional risk area; when the construction machinery is stationary, an alarm will be triggered if an unauthorized team operator enters the three-dimensional risk area.

[0053] Preferably, the system further includes

[0054] Management platform: Receives and displays risk early warning information for construction management personnel to monitor in real time; views the current status of the risk area, the movement information of construction machinery, and the identity verification results of the entering personnel, so that management personnel can quickly respond to potential risks.

[0055] The beneficial effects of the present invention are as follows: 1. Enhanced three-dimensional space recognition ability: The present invention can extract the three-dimensional space positions of construction machinery and operators from the two-dimensional images of a single camera through a monocular 3D detection algorithm. This technology solves the deficiency that it is difficult to accurately judge spatial relationships in traditional two-dimensional image detection, providing more accurate data support for risk monitoring. 2. Flexibility of dynamic risk area division: Based on the real-time position and movement state of construction machinery, the present invention automatically generates and dynamically adjusts the three-dimensional risk area around it, enabling risk monitoring to adapt to the movement of construction machinery and changes in the construction environment. Compared with statically divided risk areas, dynamic risk areas are more flexible and can more accurately adapt to complex construction scenarios. 3. Improved real-time early warning responsiveness: By judging the relative positions of personnel and construction machinery, the present invention can monitor in real time whether operators enter the risk area and trigger an early warning in a timely manner. This real-time early warning mechanism is more efficient than manual monitoring and helps to improve the safety of the construction site. 4. Cost reduction and simplified installation: The present invention uses a monocular camera for three-dimensional space detection without the need to use lidar or multi-camera systems, thus reducing equipment costs. At the same time, the installation process is simple and convenient, facilitating rapid deployment at the construction site. 5. Enhanced environmental adaptability: The present invention is applicable to complex construction environments, especially in narrow and high-density areas such as bridge construction, and can maintain stable monitoring effects. The system can adapt to changes under different lighting conditions to ensure all-weather safety monitoring. 6. Improvement of construction safety management level: The present invention pushes real-time early warning information to the management platform and relevant personnel through wireless transmission, helping construction management personnel to understand the on-site safety status in real time, facilitating quick response and decision-making, and further improving construction safety at the management level. Description of the Drawings

[0056] Figure 1 It is a flowchart of the monitoring method in the embodiment of the present invention.

[0057] Figure 2 It is a structural diagram of the monitoring system in the embodiment of the present invention. Detailed Embodiments

[0058] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0059] In order to solve the problem of difficult safety monitoring in bridge construction due to complex space and dense construction equipment, the present invention provides a method and system for monitoring dynamic three-dimensional risk areas in bridge construction based on a monocular 3D detection algorithm. Traditional two-dimensional image monitoring means cannot accurately judge the positional relationship between mechanical equipment and operators in three-dimensional space, making it difficult to effectively monitor the safety risks in the construction area. Based on the monocular 3D detection algorithm, the present invention obtains the three-dimensional positions of construction machinery and personnel from two-dimensional images through a depth estimation method, generates and dynamically adjusts the three-dimensional risk area. When an operator enters the risk area, the system automatically triggers an alarm and notifies relevant personnel in a timely manner. The present invention has the advantages of low cost, easy deployment, high real-time performance, etc., and is applicable to high-density and complex construction scenarios such as cross-sea bridges, providing a reliable technical guarantee for construction site safety management.

[0060] Embodiment 1

[0061] In this embodiment, as Figure 1 shown, the method for monitoring dynamic three-dimensional risk areas in bridge construction based on the monocular 3D detection algorithm specifically includes the following six parts:

[0062] I. Image acquisition

[0063] The video images of the construction area are collected in real time through a monocular camera. Specifically, it includes the following content:

[0064] 1.1. Select a monocular camera with high resolution and wide viewing angle to ensure coverage of the main construction area.

[0065] 1.2. Install it at a high position to avoid occlusion and ensure that the camera can monitor key construction positions within the largest range.

[0066] 1.3. The monocular camera collects the video images of the construction area in real time.

[0067] II. Image processing and 3D detection

[0068] The trained 3D detection model is used to identify the three-dimensional spatial positions and sizes of construction machinery and operators at the construction site. The 3D detection model is built using the MMDetection3D framework; the training process of the 3D model is as follows:

[0069] 2.1. Collect image data of various construction machinery and operators, and annotate the three-dimensional positions and sizes to construct a training data set for the construction scenario.

[0070] 2.2. Use ResNet101 as the backbone. First, perform multi-level feature extraction on the image data, and then pass these features to the 3D detection head to complete specific tasks. Introduce DCN to enable the backbone to more flexibly adapt to the geometric deformation of the objects in the image, thereby enhancing the effect of feature extraction.

[0071] 2.3. Combine with the 3D detection head to predict the three-dimensional information and object category of the object. The three-dimensional information includes the value of the center offset in the x direction, the value of the center offset in the y direction, depth information, the width of the object, the length of the object, the height of the object, and the rotation angle of the object.

[0072] 2.4. Adopt supervised learning technology to train the 3D detection model so that it can predict the three-dimensional spatial positions and sizes of construction machinery and operators from the video images obtained by the monocular camera.

[0073] III. Optical Flow Detection

[0074] Perform optical flow detection on the video images to analyze the motion state of the construction machinery. The specific contents are as follows.

[0075] 3.1. Input two consecutive frames of images into the LiteFlowNet2 network. The LiteFlowNet2 network estimates the optical flow information layer by layer through a cascaded convolutional neural network, estimates the pixel displacements in the entire image from coarse to fine, and outputs the optical flow field.

[0076] 3.2. Use the 2D object detection model Faster R-CNN for construction machinery object detection, output the 2D object detection bounding box, and extract the optical flow data of the target area from the optical flow field according to the 2D object detection bounding box.

[0077] 3.3. Determine whether the motion state of the construction machinery is "moving" or "stationary" by calculating the average magnitude of the optical flow vectors within the target object area.

[0078] IV. Risk Area Generation

[0079] Based on the detection results of the 3D detection model and the optical flow detection results, dynamically generate and adjust the three-dimensional risk area that moves with the construction machinery around the construction machinery. The specific contents are as follows.

[0080] 4.1. Set the risk area expansion factor, which is used to determine the size of the risk area.

[0081] 4.2. Set the value of the risk area expansion factor according to the motion state of the construction machinery. When the construction machinery is in the "moving" state, set the risk area expansion factor to 1, so that the size of the generated risk area is equivalent to 1 times the distance of the construction machinery's bounding box. When the construction machinery is in the "stationary" state, set the risk area expansion factor to 0.5, so that the size of the generated risk area is reduced to 0.5 times the distance of the construction machinery's bounding box.

[0082] 4.3. Extract the basic tensor data of the construction machinery's bounding box, including position coordinates, length, and width.

[0083] 4.4. Use the risk area expansion factor to expand the specified tensor data. Only adjust the length and width parameters of the construction machinery's bounding box proportionally, and generate a new three-dimensional risk area bounding box using the adjusted tensor data. Specifically, multiply the length and width in the tensor data by (1 + expansion factor) to achieve the expansion operation, while the height remains unchanged.

[0084] In this embodiment, the risk area can move automatically as the construction machinery moves, and there is no need for manual re-delineation of the risk area.

[0085] V. Identity Recognition

[0086] Real-time identify the identity of the personnel entering the risk area by comparing face recognition with the identity database to confirm whether they are authorized operators of the construction machinery operation team. The specific contents are as follows.

[0087] 5.1. Collect and register the facial features of the operation team members, and generate feature codes to be stored in the identity database.

[0088] 5.2. Real-time capture the images of the personnel entering the risk area.

[0089] 5.3. Use the face detection model and face recognition model to identify and extract the face feature vectors of the personnel images, compare the face feature vectors of the personnel with all the feature vectors in the identity database, and calculate the similarity between the two to determine whether the corresponding personnel are authorized personnel.

[0090] VI. Early Warning

[0091] Real-time monitor the positions of the operators. When the construction machinery is running, an alarm will be triggered if a person other than the operator enters the three-dimensional risk area. When the construction machinery is stationary, an alarm will be triggered if a person other than an authorized team member enters the three-dimensional risk area.

[0092] In this embodiment, as Figure 2 shown, a bridge construction dynamic three-dimensional risk area monitoring system based on a monocular 3D detection algorithm is provided. The system can implement the above-mentioned method. The system includes the following seven modules.

[0093] I. Image Acquisition Module

[0094] 1.1 Equipment Selection: Select a monocular camera with high resolution and wide viewing angle to ensure coverage of the main construction area.

[0095] 1.2 Installation Location: Install it at a high position to avoid occlusion and ensure that the camera can monitor key construction positions within the maximum range.

[0096] 1.3 Real-time Transmission: Use the network to transmit the acquired images to the edge computing device in real time for 3D detection and risk analysis.

[0097] In this embodiment, a monocular camera is combined with a 3D detection model to achieve three-dimensional spatial positioning of construction machinery and personnel. Compared with traditional multi-sensor solutions, the monocular solution significantly reduces the hardware cost and deployment complexity, is suitable for large-scale applications, and meets the requirements of precise on-site positioning.

[0098] II. Image Processing and 3D Detection Module

[0099] Based on the MMDetect ion3D framework, establish a 3D detection model for construction machinery and operators to identify the three-dimensional spatial positions and dimensions of construction machinery and operators at the construction site.

[0100] 2.1 Model Training:

[0101] (1) Data Preparation: Collect image data of various construction machinery and operators, and annotate the three-dimensional positions and dimensions to construct a training data set for the construction scenario.

[0102] (2) Depth Estimation: Use ResNet101 as the backbone. First, perform multi-level feature extraction on the image, and then pass these features to other modules (such as the detection head) to complete specific tasks. Introducing DCN (Deformable Convolution) enables this backbone to more flexibly adapt to the geometric deformations of objects in the image, thereby enhancing the effect of feature extraction.

[0103] (3) Combine with a 3D detection head (FCOS3D) to predict the three-dimensional stereo information and object category of the object. The three-dimensional stereo information includes Δx: the value of the center offset in the x direction. Δy: the value of the center offset in the y direction. d: depth information. w: the width of the object. l: the length of the object. h: the height of the object. θ: the rotation angle of the object.

[0104] (4) Training Process: Use supervised learning technology to train the model so that it can predict the three-dimensional positions and dimensions of construction machinery and operators from monocular images.

[0105] 2.2. Real-time Inference:

[0106] (1) Input Data: Real-time receive camera image data and input it into the trained 3D detection model.

[0107] (2) Output Results: The model outputs the categories, 3D positions, and dimensions of construction machinery and operators, providing basic data for the risk area generation module.

[0108] III. Optical Flow Detection Module

[0109] Introduce an optical flow module to analyze the motion state of construction machinery and adjust the size of the risk area according to the state.

[0110] 3.1. Optical Flow Calculation: Input two consecutive frames of images into LiteFlowNet2 to calculate the optical flow field.

[0111] Select the LiteFlowNet2 network to process the image pair to obtain the optical flow field. The LiteFlowNet2 network is an efficient optical flow estimation algorithm, and the main steps are as follows.

[0112] The image pair is two consecutive frames of images (frame_t, frame_(t + 1)), representing the scenes at a certain moment t and the next moment t + 1.

[0113] (1) Input to the Network: Input the two frames of images into the LiteFlowNet2 network. The network estimates the optical flow information layer by layer through a cascaded convolutional neural network, estimating the pixel displacements in the entire image from coarse to fine.

[0114] (2) Obtain the Optical Flow Field: The output is an optical flow field, which is a tensor with a shape of (H, W, 2), where H and W are the height and width of the image, and 2 represents the horizontal (dx) and vertical (dy) displacements of each pixel. The output optical flow field can be understood as the displacement vector of each pixel point between the two frames of images.

[0115] 3.2. Extract Optical Flow Information of the Target Area: Use the 2D detection model Faster R-CNN to perform construction machinery target detection, output the 2D target detection bounding box, and extract the optical flow data of the target object area from the optical flow field.

[0116] (1) Use a 2D target detection model (Faster R-CNN) to detect construction machinery in the image.

[0117] (2) The target detection model provides a 2D bounding box for the construction machinery, that is, the position of the construction machinery in the image.

[0118] (3) According to the bounding box of the 2D target detection, extract the optical flow information of the target area from the optical flow field.

[0119] 3.3. Determine the motion state: By calculating the average magnitude of the optical flow vectors within the target area, determine whether the object is in motion. The optical flow module determines the state of the construction machinery as "moving" or "stationary".

[0120] In this embodiment, the motion state of the construction machinery is utilized in real-time by the optical flow module to intelligently adjust the expansion factor of the risk area. This dynamic adjustment method enables the system to adapt to the moving or stationary state of the construction machinery, improves the environmental adaptability, and ensures the real-time and accuracy of the risk area.

[0121] IV. Generation of Risk Area

[0122] Combined with the construction machinery state information output by the optical flow detection module, the generation and adjustment of the dynamic risk area are realized.

[0123] 4.1. Setting of the risk area expansion factor: Set a risk area expansion factor, which is used to determine the size of the risk area.

[0124] 4.2. Dynamic setting of the risk area expansion factor: According to the judgment result of the construction machinery motion state, set the value of the risk area expansion factor.

[0125] When the construction machinery is in the "moving" state, set the risk area expansion factor to 1, so that the generated risk area size is equivalent to 1 times the distance of the construction machinery bounding box.

[0126] When the construction machinery is in the "stationary" state, set the risk area expansion factor to 0.5, so that the generated risk area size is reduced to 0.5 times the distance of the construction machinery bounding box.

[0127] 4.3. Acquisition and preprocessing of bounding box data: Extract the basic tensor data of the 3D bounding box (construction machinery bounding box) and generate a copy of it. Include position coordinates (x, y, z), length, and width.

[0128] 4.4. Risk area expansion operation: On the generated 3D bounding box tensor, use the risk area expansion factor to expand the specified parameters.

[0129] Only perform proportional adjustment on the length and width parameters of the bounding box to form the expanded risk area bounding box. Specifically, the expansion operation is achieved by multiplying the length and width of the tensor by (1 + expansion factor) respectively, while the height parameter remains unchanged to ensure the three-dimensional structure rationality of the risk area.

[0130] Generate a new three-dimensional risk area bounding box object using the above-expanded tensor data.

[0131] In this embodiment, based on the three-dimensional position and motion state of construction machinery, the system realizes automatic follow-up adjustment of the dynamic risk area. Different from the traditional 2D risk area, since the 2D image lacks depth information, the risk area usually needs to be manually delimited. The 3D risk area can automatically move with the vehicle or construction machinery, ensuring that the risk area always accurately covers the high-risk area, greatly improving the automation and accuracy of risk control.

[0132] V. Identity Recognition Module

[0133] 5.1 Face Detection Module

[0134] (1) Function: Identify the face area in the video stream or image.

[0135] (2) Technology Selection: Use a deep learning-based face detection model, MTCNN (Multi-task Cascaded Convolutional Networks).

[0136] (3) Process: Input the real-time image frame, and input the frame data into the face detection model. The model outputs the bounding box coordinates of the face for subsequent extraction of face features.

[0137] 5.2 Face Feature Extraction Module

[0138] (1) Function: Encode the detected face area to generate a feature vector that can be used for identity comparison.

[0139] (2) Technology Selection: Use a pre-trained face recognition model (ArcFace) to extract the feature vector of each face.

[0140] (3) Process: Intercept the detected face area and input it into the feature extraction model. The model outputs a fixed-length feature vector (usually 128 or 512 dimensions), which represents the unique feature information of the face. The feature vector is used for comparison in the identity database.

[0141] 5.3 Identity Database Construction and Management Module

[0142] (1) Function: Establish and manage the identity database of the operation team members. The database contains the face feature encodings of each authorized operator.

[0143] (2) Data Collection: Before the system starts, collect multiple face images of each operation team member.

[0144] (3) Feature Encoding: Pass each face image through the feature extraction module, generate a feature vector and store it in the identity database.

[0145] (4) Database Management: Provide functions for adding, deleting, modifying, and querying the identity database to ensure real-time updates of authorized personnel information.

[0146] 5.4 Identity Comparison and Authorization Verification Module

[0147] (1) Function: Authenticate the identity of personnel entering the area to determine whether they are authorized personnel.

[0148] (2) Comparison Method: Calculate the similarity (Euclidean distance) between the newly detected face feature vector and each feature vector in the identity database. Set a similarity threshold (e.g., 0.8). When the similarity is higher than the threshold, the person is determined to be an authorized person; otherwise, they are an unauthorized person.

[0149] (3) Process: When a face entering the area is detected, extract its feature vector. Compare the feature vector with all feature vectors in the identity database and calculate the similarity. If the highest similarity exceeds the preset threshold, confirm the identity of the person as an authorized operator; if the condition is not met, mark them as an unauthorized person.

[0150] (4) Output: According to the comparison result, the control system performs different operations:

[0151] (5) Authorized Personnel: Are allowed to enter the risk area, and a "Allowed to Enter" prompt is displayed. Unauthorized Personnel: Trigger the warning system and are prohibited from entering the risk area.

[0152] 5.5 System Workflow

[0153] (1) Initialization: Collect and register the facial features of the operating team personnel, generate feature codes, and store them in the identity database.

[0154] (2) Real-time Monitoring: During the operation of the system, capture the images of personnel entering the risk area in real time.

[0155] (3) Identity Verification: Use the face detection and feature extraction module to identify and extract the face feature vector of the new person. Compare the new feature with the identity database, calculate the similarity, and determine whether they are authorized personnel.

[0156] VI. Warning Module

[0157] Real-time monitor the positions of the operating personnel. When the construction machinery is running, an alarm will be triggered if a person who is not the operator enters the three-dimensional risk area; when the construction machinery is stationary, an alarm will be triggered if an unauthorized team operating personnel enters the three-dimensional risk area.

[0158] 6.1 Relative Position Calculation: Use the data of the 3D detection module to calculate the spatial relative distance between the operating personnel and the construction machinery in real time.

[0159] Monitor the location of operators in real time. When the construction machinery is running, if a non-operator driver enters the risk area, an alarm will be triggered; when the construction machinery is stationary, if an unauthorized person from the operation team enters the risk area, an alarm will be triggered to notify relevant personnel.

[0160] 6.2. Alarm mechanism: When the system detects that the alarm condition is triggered, it immediately triggers an audible and visual alarm to alert on-site operators.

[0161] Send the alarm information to the management platform and the mobile devices of relevant management personnel through the wireless communication module to ensure that the early warning information is received in a timely manner.

[0162] In this embodiment, a multi-level intelligent early warning mechanism combining 3D detection, dynamic risk area division, and optical flow analysis is provided. The system automatically judges the risk level and triggers different levels of early warnings, including on-site audible and visual alarms, remote notification of management personnel, real-time recording of risk events, etc. Through multi-level response and closed-loop control, it ensures that the early warning signal is synchronized with the actual risk situation, effectively improving the safety management efficiency and intelligent level of the construction site.

[0163] VII. Management Platform

[0164] Receive and display risk early warning information for construction management personnel to monitor in real time; view the current status of the risk area, the movement information of construction machinery, and the identity verification results of the entering personnel, so that management personnel can quickly respond to potential risks.

[0165] 7.1. Platform construction:

[0166] Design a risk monitoring interface on the management platform to display the locations of construction machinery and operators, the status of the risk area, and early warning information in real time.

[0167] The platform receives early warning information from front-end devices and saves the alarm records for later analysis.

[0168] 7.2. Function implementation:

[0169] (1) Real-time monitoring: Dynamically display construction machinery, operators, and risk areas in a 3D scene.

[0170] (2) Early warning record: Save the time, location, and construction machinery information when the early warning is triggered for tracking and review.

[0171] (3) Remote control: Management personnel can adjust system parameters (such as safety distance) through the platform to meet the requirements of different construction scenarios.

[0172] By adopting the above technical solutions disclosed in the present invention, the following beneficial effects are obtained:

[0173] The present invention provides a method and system for monitoring dynamic three-dimensional risk areas in bridge construction based on a monocular 3D detection algorithm, with enhanced three-dimensional space recognition ability: Through the monocular 3D detection algorithm, the present invention can extract the three-dimensional spatial positions of construction machinery and operators from the two-dimensional images of a single camera. This technology solves the deficiency that it is difficult to accurately judge spatial relationships in traditional two-dimensional image detection, providing more accurate data support for risk monitoring. Flexibility in dynamic risk area division: Based on the real-time position and movement state of construction machinery, the present invention automatically generates and dynamically adjusts the three-dimensional risk areas around it, enabling risk monitoring to adapt to the movement of equipment and changes in the construction environment. Compared with statically divided risk areas, dynamic risk areas are more flexible and can more precisely adapt to complex construction scenarios. Improved real-time warning responsiveness: By judging the relative positions of personnel and construction machinery, the present invention can monitor in real time whether operators enter the risk area and trigger warnings in a timely manner. This real-time warning mechanism is more efficient than manual monitoring and helps to improve the safety of the construction site. Cost reduction and simplified installation: The present invention uses a monocular camera for three-dimensional space detection without the need to use lidar or multi-camera systems, thus reducing equipment costs. At the same time, the installation process is simple and convenient, facilitating rapid deployment at the construction site. Enhanced environmental adaptability: The present invention is applicable to complex construction environments, especially in narrow and high-density areas such as bridge construction, and can maintain stable monitoring effects. The system can adapt to changes under different lighting conditions to ensure all-weather safety monitoring. Improvement of construction safety management level: The present invention pushes real-time warning information to the management platform and relevant personnel through wireless transmission, helping construction management personnel to understand the on-site safety status in real time, facilitating quick response and decision-making, and further improving construction safety from the management level.

[0174] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A monitoring method for the dynamic three-dimensional risk area of bridge construction based on a monocular 3D detection algorithm, characterized in that: It includes the following steps: S1. Image acquisition: Real-time video images of the construction area are acquired through a monocular camera. S2. Image processing and 3D detection: The trained 3D detection model is used to identify the three-dimensional spatial positions and sizes of construction machinery and operators at the construction site. S3. Optical flow detection: Optical flow detection is performed on the video images to analyze the motion state of the construction machinery. S4. Risk area generation: Based on the detection results of the 3D detection model and the optical flow detection results, a three-dimensional risk area that moves with the construction machinery around the construction machinery is dynamically generated and adjusted. S5. Identity recognition: By comparing face recognition with the identity database, the identity of the personnel entering the risk area is recognized in real time to confirm whether they are authorized operators of the construction machinery operation team. S6. Warning: The positions of the operators are monitored in real time. When the construction machinery is running, an alarm will be triggered if a person other than the operator enters the three-dimensional risk area; when the construction machinery is stationary, an alarm will be triggered if a person other than the authorized team operator enters the three-dimensional risk area.

2. The monitoring method for the dynamic three-dimensional risk area of bridge construction based on the monocular 3D detection algorithm according to claim 1, wherein: Step S1 specifically includes the following content: S11. Select a monocular camera with high resolution and wide viewing angle to ensure coverage of the main construction area. S12. Install it at a high position to avoid occlusion and ensure that the camera can monitor the key construction positions within the maximum range. S13. The monocular camera real-time acquires video images of the construction area.

3. The monitoring method for the dynamic three-dimensional risk area of bridge construction based on the monocular 3D detection algorithm according to claim 1, wherein: The 3D detection model is built using the MMDetection3D framework; the training process of the 3D model is as follows: S21. Image data of various construction machinery and operators are acquired, and the three-dimensional positions and sizes are marked to construct a training data set for the construction scene. S22. Use ResNet101 as the backbone. First, multi-level feature extraction is performed on the image data, and then these features are passed to the 3D detection head to complete specific tasks; introducing DCN enables this backbone to more flexibly adapt to the geometric deformation of the targets in the image, thereby enhancing the effect of feature extraction. S23. Combine with the 3D detection head to predict the three-dimensional stereo information and object categories of the objects; the three-dimensional stereo information includes the value of the center offset in the x direction, the value of the center offset in the y direction, depth information, the width of the object, the length of the object, the height of the object, and the rotation angle of the object. S24. Use supervised learning technology to train the 3D detection model so that it can predict the three-dimensional spatial positions and sizes of construction machinery and operators from the video images obtained by the monocular camera.

4. The monitoring method for the dynamic three-dimensional risk area of bridge construction based on the monocular 3D detection algorithm according to claim 1, wherein: Step S3 specifically includes the following content: S31. Input two consecutive frames of images into the LiteFlowNet2 network. The LiteFlowNet2 network estimates the optical flow information layer by layer through a cascaded convolutional neural network, estimates the pixel displacements in the entire image from coarse to fine, and outputs the optical flow field. S32. Use the 2D object detection model Faster R-CNN for construction machinery object detection, output the 2D object detection bounding box, and extract the optical flow data of the target area from the optical flow field according to the 2D object detection bounding box. S33. Determine whether the motion state of the construction machinery is "moving" or "stationary" by calculating the average magnitude of the optical flow vectors within the target object area.

5. The monitoring method for the dynamic three-dimensional risk area of bridge construction based on the monocular 3D detection algorithm according to claim 1, wherein: Step S4 specifically includes the following content. S41. Set a risk area expansion factor, which is used to determine the size of the risk area. S42. Set the value of the risk area expansion factor according to the motion state of the construction machinery. S43. Extract the basic tensor data of the construction machinery bounding box, including position coordinates, length, and width. S44. Use the risk area expansion factor to expand the specified tensor data.

6. The method for monitoring the dynamic three-dimensional risk area of bridge construction based on the monocular 3D detection algorithm according to claim 5, wherein: Specifically for step S42, when the construction machinery is in the "moving" state, set the risk area expansion factor to 1, so that the generated risk area size is equivalent to 1 times the distance of the construction machinery bounding box; when the construction machinery is in the "stationary" state, set the risk area expansion factor to 0.5, so that the generated risk area size is reduced to 0.5 times the distance of the construction machinery bounding box.

7. The method for monitoring the dynamic three-dimensional risk area of bridge construction based on the monocular 3D detection algorithm according to claim 5, wherein: Specifically for step S44, only proportionally adjust the length and width parameters of the construction machinery bounding box, and generate a new three-dimensional risk area bounding box using the adjusted tensor data. Multiply the length and width in the tensor data by (1 + expansion factor) respectively to achieve the expansion operation, while keeping the height unchanged.

8. The method for monitoring the dynamic three-dimensional risk area of bridge construction based on the monocular 3D detection algorithm according to claim 1, wherein: Step S5 specifically includes the following content. S51. Collect and register the facial features of the operating team workers, generate feature codes and store them in the identity database. S52. Real-time capture the images of the personnel entering the risk area. S53. Use the face detection model and face recognition model to identify and extract the face feature vectors of the personnel images, compare the face feature vectors of the personnel with all the feature vectors in the identity database, calculate the similarity between the two, and determine whether the corresponding personnel are authorized personnel.

9. A bridge construction dynamic three-dimensional risk area monitoring system based on a monocular 3D detection algorithm, characterized in that: Including Image acquisition module: Real-time collect the video images of the construction area through a monocular camera. Image processing and 3D detection module: Use the trained 3D detection model to identify the three-dimensional spatial positions and sizes of the construction machinery and operating personnel at the construction site. Optical flow detection module: Perform optical flow detection on the video images to analyze the motion state of the construction machinery. Risk area generation: Dynamically generate a three-dimensional risk area around the construction machinery based on the detection results of the 3D detection model and the optical flow detection results. Identity recognition module: Real-time identify the identity of the personnel entering the risk area through face recognition and comparison with the identity database, and confirm whether they are authorized operating personnel of the construction machinery operation team. Early warning module: Real-time monitor the positions of the operating personnel. When the construction machinery is running, an alarm will be triggered if a person other than the operator enters the three-dimensional risk area; when the construction machinery is stationary, an alarm will be triggered if a person other than the authorized team operating personnel enters the three-dimensional risk area.

10. The bridge construction dynamic three-dimensional risk area monitoring system based on the monocular 3D detection algorithm according to claim 9, characterized in that: The system also includes Management platform: Receive and display the risk warning information for the construction management personnel to monitor in real time; view the current risk area status, construction machinery motion information, and identity verification results of the entering personnel, so that the management personnel can quickly respond to potential risks.

Images