Intelligent monitoring method and system for the posture of precast concrete components

Through binocular vision system and object detection model, circular targets are monitored and identified in real time, and three-dimensional coordinate reconstruction is carried out, which solves the problems of low efficiency and large errors in traditional installation, and realizes efficient and accurate posture monitoring of precast concrete components and Installation instructions.

CN118799811BActive Publication Date: 2024-11-15HUNAN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411259494.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-10
Publication Date
2024-11-15
Estimated Expiration
2044-09-10

AI Technical Summary

Technical Problem

During the installation of traditional precast concrete components, relying on the theodolite and mirrors, the installation quality depends on marking and measurement errors, resulting in low efficiency and time-consuming and requires a lot of human resources.

Method used

The binocular vision system is used to combine a pre-trained object detection model, and the circular target is monitored and identified in real time through the binocular camera and preset elliptical center detection method, and three-dimensional world coordinate reconstruction is carried out to construct the pose information of the components, and guide the installation process.

Benefits of technology

It realizes real-time and accurate monitoring of the position information of prefabricated lifting components under the conditions of improving efficiency, reducing the waste of human resources and improving installation efficiency and accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118799811B_ABST
    Figure CN118799811B_ABST
Patent Text Reader

Abstract

The present application discloses a method and system for intelligent monitoring of the posture of precast concrete components, which relates to the field of target detection, including: in the current assembly scene of hoisting bridge components, through a binocular vision system, real-time monitoring of the hoisting site where a circular target has been pre-installed on the precast hoisting component and the target monitoring position, and target recognition to obtain monitoring results; when the monitoring results show that the circular target on the precast hoisting component is currently identified, the center pixel coordinates of the identified circular target and the target center pixel coordinates are extracted; based on the target center pixel coordinates, the three-dimensional world coordinates of the target center point are reconstructed to obtain the target three-dimensional world coordinates; by using the target three-dimensional world coordinates to construct a component coordinate system and a reference coordinate system representing the current state and the design state respectively, the current posture information of the precast hoisting component relative to the design state is obtained, so as to guide the installation of the component based on the posture information. It can effectively improve efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of target detection, and in particular to a method and system for intelligent monitoring of the posture of precast concrete components. Background Art

[0002] Prefabrication creates a cost-effective structure by quickly assembling prefabricated hoisting components. It effectively speeds up bridge construction and reduces traffic disruption. The traditional hoisting assembly process uses reflectors or theodolites and total stations to assist in manual monitoring of hoisting components, which means that the crane operator on site only relies on the gestures or voices of other workers. Then, the two theodolites work together to control vertical tilt and horizontal deviation. Adjustments will be made based on the readings of the theodolite, and sometimes repeated re-lifting and lowering may be required, which is time-consuming and slows down the construction progress.

[0003] To solve the above problems, the existing technical solution is to lift the prefabricated pier to the designed position by a crawler crane, and evaluate the installation quality by two theodolites for guidance. The two theodolites monitor the two adjacent surfaces of the prefabricated pier together. Figure 1 As shown, two theodolites are configured to be located on two horizontal principal axes. For each monitoring surface, two circular gauge marks with crosshairs and vertical scale lines are pasted along the vertical centerline of the pier surface. In an ideal installation state, only the vertical angle of the theodolite needs to be changed, and the crosshairs of the theodolite are aligned with the crosshairs on the gauge mark. The installation process includes two main steps. Step 1: When the pier is close to the steel bar on the pier cap, horizontal adjustments are made so that the steel bar is inserted into the sleeve cavity. This is guided by observing from the reflector at the bottom and the reading of the theodolite. Step 2: After the pier falls down to the support point, a hydraulic jack is used to make fine adjustments in the vertical direction to ensure that the pier is in an upright state. This step is guided only by the reading of the theodolite.

[0004] As a result, the installation quality is highly dependent on the configuration of the markers and theodolite, which has a certain degree of error. And because the horizontal error measurement capability of the theodolite is limited, repeated horizontal alignment work may be required before a satisfactory vertical deviation reading of the theodolite is obtained, which is not only inefficient but also consumes a lot of human resources. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a method and system for intelligent monitoring of the posture of precast concrete components, which can effectively monitor the posture information of precast hoisting components in real time and accurately to provide guidance for component installation while improving efficiency, and avoid unnecessary waste of human resources. The specific scheme is as follows:

[0006] In a first aspect, the present application provides a method for intelligently monitoring the posture of a precast concrete component, comprising:

[0007] In the current hoisting bridge component assembly scenario, a binocular vision system in a tripod at a preset distance from the prefabricated hoisting component is used to monitor the hoisting site where a circular target has been preinstalled at the prefabricated hoisting component and the target monitoring position in real time, and target recognition is performed on the monitoring image to obtain corresponding monitoring results; wherein the binocular vision system is a system equipped with a binocular camera and a pre-trained target detection model;

[0008] When the monitoring result indicates that the circular target on the prefabricated hoisting component is currently identified, the central pixel coordinates of the identified circular target are extracted based on a preset ellipse center detection method to obtain corresponding target central pixel coordinates;

[0009] Reconstructing the three-dimensional world coordinates of the target center point based on the three-dimensional measurement principle of binocular vision and the target center pixel coordinates to obtain the corresponding target three-dimensional world coordinates;

[0010] By using the target three-dimensional world coordinates to construct a component coordinate system and a reference coordinate system representing the current state and the design state respectively, the current posture information of the prefabricated hoisting component relative to the design state is obtained, so as to guide the installation of the component based on the posture information.

[0011] Optionally, the method further includes:

[0012] A total station is used to pre-install pre-configured circular targets at the monitorable locations of the prefabricated hanging components and the designed reference positions.

[0013] Optionally, before the binocular vision system in a tripod at a preset distance from the prefabricated hoisting component is used to perform real-time monitoring on the hoisting site where the circular target has been preinstalled on the prefabricated hoisting component and the target monitoring position, the method further includes:

[0014] Build a dataset containing multiple images;

[0015] After cropping each of the images in the data set to a uniform size, dividing the data set to obtain a corresponding training data set and a verification data set;

[0016] Based on the training data set and the verification data set, the constructed original YOLO-ODCA network is trained and verified respectively to obtain a trained target detection model; wherein the target detection model includes a YOLOv7 network, a full-dimensional dynamic convolution module, a coordinate attention module, and a normalized distance module.

[0017] Optionally, before respectively training and verifying the constructed original YOLO-ODCA network based on the training data set and the verification data set, the method further includes:

[0018] In the process of constructing the original YOLO-ODCA network, the full-dimensional dynamic convolution is introduced into the efficient layer aggregation network in YOLOv7, the coordinate attention mechanism is introduced into the last concat structure of the sampling structure on the neck network, the normalized Gaussian Wasserstein distance is used instead of the original IoU as the metric, and NWD-Loss is introduced on the basis of CIoU-Loss to obtain the original YOLO-ODCA network.

[0019] Optionally, when the monitoring result indicates that the circular target on the prefabricated hoisting component is currently identified, extracting the central pixel coordinates of the identified circular target based on a preset ellipse center detection method to obtain the corresponding target central pixel coordinates includes:

[0020] When the monitoring result indicates that the circular target on the prefabricated hanging component is currently identified, the area where the circular target is located is cropped out from the image according to the pixel coordinates of the current detection frame to obtain a corresponding area to be detected;

[0021] Based on a preset ellipse center detection method, an ellipse center detection is performed on the area to be detected to obtain target center pixel coordinates corresponding to the identified circular target.

[0022] Optionally, the three-dimensional measurement principle based on binocular vision and the target center pixel coordinates are used to reconstruct the three-dimensional world coordinates of the target center point to obtain the corresponding three-dimensional world coordinates of the target, including:

[0023] By calibrating the binocular camera, an intrinsic parameter matrix and an extrinsic parameter matrix of each of a left camera and a right camera in the binocular camera, and a position conversion coefficient between the left camera and the right camera are obtained; wherein the extrinsic parameter matrix includes a rotation matrix and a displacement matrix for representing the conversion of coordinates in a world coordinate system to a camera coordinate system;

[0024] Based on the three-dimensional measurement principle of binocular vision, rays are drawn using the target center pixel coordinates, the intrinsic parameter matrix, the extrinsic parameter matrix, and the position conversion coefficient, and the intersection of the rays in the space coordinate system is used as the target three-dimensional world coordinates.

[0025] Optionally, the step of constructing a component coordinate system and a reference coordinate system respectively representing a current state and a design state by using the target three-dimensional world coordinates to obtain the current posture information of the prefabricated hoisting component relative to the design state includes:

[0026] A component coordinate system representing the current state is constructed on the prefabricated hoisting component using the target three-dimensional world coordinates, and a reference coordinate system representing the design state is constructed on the existing structure; wherein the existing structure is a structure that has been installed in the current hoisting bridge component assembly scene and is designed to install the prefabricated hoisting component;

[0027] Selecting a corresponding target control point according to the structural design information, and calculating target coordinate information of the target control point in the component coordinate system;

[0028] The target coordinate information is converted in the reference coordinate system and deviation is calculated to obtain the current posture information of the prefabricated hoisting component relative to the design state, and the posture information is output online; wherein the posture information includes corresponding vertical deviation information and horizontal deviation information.

[0029] In a second aspect, the present application provides a precast concrete component posture intelligent monitoring system, comprising:

[0030] A real-time monitoring module is used to perform real-time monitoring of the hoisting site where circular targets have been pre-installed at the prefabricated hoisting components and target monitoring positions through a binocular vision system in a tripod at a preset distance from the prefabricated hoisting components in the current hoisting bridge component assembly scene, and to perform target recognition on the monitoring image to obtain corresponding monitoring results; wherein the binocular vision system is a system equipped with a binocular camera and a pre-trained target detection model;

[0031] A center coordinate acquisition module, for extracting the center pixel coordinates of the identified circular target based on a preset ellipse center detection method to obtain the corresponding target center pixel coordinates when the monitoring result indicates that the circular target on the prefabricated hoisting component is currently identified;

[0032] A three-dimensional coordinate acquisition module is used to reconstruct the three-dimensional world coordinates of the target center point based on the three-dimensional measurement principle of binocular vision and the target center pixel coordinates to obtain the corresponding target three-dimensional world coordinates;

[0033] The posture information acquisition module is used to construct a component coordinate system and a reference coordinate system representing the current state and the design state respectively by using the target three-dimensional world coordinates, so as to obtain the current posture information of the prefabricated lifting component relative to the design state, so as to guide the installation of the component based on the posture information.

[0034] It can be seen that in the present application, in the current lifting bridge component assembly scene, a binocular vision system in a tripod at a preset distance from the prefabricated lifting component is used to monitor the lifting site where a circular target has been preinstalled on the prefabricated lifting component and the target monitoring position in real time, and target recognition is performed on the monitoring image to obtain corresponding monitoring results; wherein the binocular vision system is a system equipped with a binocular camera and a pre-trained target detection model; when the monitoring result shows that the circular target on the prefabricated lifting component is currently identified, the center pixel coordinates of the identified circular target are extracted based on a preset ellipse center detection method to obtain the corresponding target center pixel coordinates; the three-dimensional world coordinates of the target center point are reconstructed based on the three-dimensional measurement principle of binocular vision and the target center pixel coordinates to obtain the corresponding target three-dimensional world coordinates; by using the target three-dimensional world coordinates to construct a component coordinate system and a reference coordinate system representing the current state and the design state respectively, the current posture information of the prefabricated lifting component relative to the design state is obtained, so as to guide the installation of the component based on the posture information. That is, in this application, a binocular vision system equipped with a binocular camera and a pre-trained target detection model monitors the hoisting site in real time, and when the circular target pre-installed on the prefabricated hoisting component is detected, the target center pixel coordinates are extracted based on the preset ellipse center detection method, and the corresponding target three-dimensional world coordinates are obtained based on the three-dimensional measurement principle of binocular vision, so as to obtain the current posture information of the prefabricated hoisting component relative to the design state, so as to guide the installation of the component based on the posture information. In this way, the posture information of the prefabricated hoisting component can be effectively monitored in real time and accurately to guide the installation of the component while improving efficiency, and avoid unnecessary waste of human resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0036] Figure 1 A monitoring schematic diagram of a traditional monitoring scheme for a prefabricated bridge pier hoisting site provided in this application;

[0037] Figure 2 A flow chart of a method for intelligent monitoring of the posture of precast concrete components provided in this application;

[0038] Figure 3 A schematic diagram of a target pre-installation provided in this application;

[0039] Figure 4 A schematic diagram of a binocular vision-based hoisting component posture monitoring framework provided in this application;

[0040] Figure 5 A schematic diagram of constructing a circular target data set provided in this application;

[0041] Figure 6 A schematic diagram of circular target center detection provided in this application;

[0042] Figure 7 A schematic diagram of a reprojection error provided in this application;

[0043] Figure 8 A schematic diagram of a triangulation model based on binocular vision provided in this application;

[0044] Fig. 9 A schematic diagram of a hoisting component posture monitoring based on binocular vision provided in this application;

[0045] Fig.10 A schematic diagram of a specific binocular vision-based hoisting component posture monitoring framework provided in this application;

[0046] Fig.11 A schematic diagram of the structure of an intelligent monitoring system for the posture of precast concrete components provided in this application. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] In the existing technical solution, the prefabricated pier is lifted to the designed position by a crawler crane, and the installation quality is evaluated by two theodolites for guidance. The two theodolites monitor the two adjacent surfaces of the prefabricated pier together. Figure 1As shown, two theodolites are configured to be located on two horizontal main axes. For each monitoring surface, two circular gauge marks with crosshairs and vertical scale lines are pasted along the vertical centerline of the pier surface. In an ideal installation state, the crosshairs of the theodolite are aligned with the crosshairs on the gauge mark by simply changing the vertical angle of the theodolite. The installation process includes two main steps. Step 1: When the pier is close to the steel bars on the pier cap, horizontal adjustments are made so that the steel bars are inserted into the sleeve cavity. This is guided by observing the reflector from the bottom and the reading of the theodolite. Step 2: After the pier falls down to the support point, a hydraulic jack is used to make vertical fine adjustments to ensure that the pier is in an upright state. This step is guided only by the reading of the theodolite. In this way, the installation quality is highly dependent on the configuration of the marks and theodolite, and there is a certain error. And because the horizontal error measurement capability of the theodolite is limited. Therefore, repeated horizontal alignment work may be required before a satisfactory vertical deviation reading measured by the theodolite is obtained, which is not only inefficient but also consumes a lot of human resources.

[0049] To this end, the present application provides a precast concrete bridge component posture monitoring solution, which can effectively monitor the posture information of precast hoisting components in real time and accurately while improving efficiency to provide component installation guidance and avoid unnecessary waste of human resources.

[0050] See also Figure 2 As shown, an embodiment of the present invention discloses a method for intelligently monitoring the posture of a precast concrete component, comprising:

[0051] Step S11: In the current lifting bridge component assembly scene, a binocular vision system in a tripod at a preset distance from the prefabricated lifting component is used to perform real-time monitoring of the lifting site where circular targets have been preinstalled on the prefabricated lifting components and the target monitoring position, and target recognition is performed on the monitoring image to obtain corresponding monitoring results; wherein the binocular vision system is a system equipped with a binocular camera and a pre-trained target detection model.

[0052] In this embodiment, in the current hoisting bridge component assembly scenario, before monitoring, a total station is first used to pre-install pre-configured circular targets at the monitorable positions of the prefabricated hoisting components and the design reference positions to ensure that the lines connecting the centers of the two circular targets constituting the two axes of the coordinate system are perpendicular to each other. Figure 3 The circular targets (i.e., artificial targets) are composed of 6 circular targets made of A4 paper, such as Figure 3 As shown in the figure, the target consists of a large circle (50 mm in diameter) in the middle and small circles (20 mm in diameter) around it. The large circle is used to measure the coordinates of the center, and the small circles are used to distinguish artificial targets by numbering the targets.

[0053] It is understandable that, since the monitoring process cannot ensure that the binocular camera always shoots the target vertically, the circular target will become an elliptical target in the monitoring image.

[0054] In a specific embodiment, after the target is pre-installed, a binocular vision system equipped with a binocular camera with a focal length of 12 mm is arranged at a distance of 10 m from the prefabricated hoisting component to monitor the hoisting site of the pre-installed circular target in real time. The elliptical target in the monitoring image is recognized and classified in real time by the proposed improved target detection model. The stereoscopic vision system is set on a tripod in the front and can be used to observe the prefabricated hoisting component and the existing structure, wherein the existing structure is a structure that has been installed in the current hoisting bridge component assembly scene and is designed to install the prefabricated hoisting component.

[0055] Combination Figure 4 As shown, regarding the target detection algorithm, before performing elliptical target detection, it is necessary to construct a data set of elliptical targets for training the improved target detection algorithm based on deep learning. That is, before the binocular vision system in the tripod at a preset distance from the prefabricated hoisting component is used to conduct real-time monitoring of the hoisting site where the circular target has been preinstalled on the prefabricated hoisting component and the target monitoring position, it also includes: constructing a data set containing multiple images; after uniformly cropping each of the images in the data set, dividing the data set to obtain a corresponding training data set and a verification data set; based on the training data set and the verification data set, respectively, the constructed original YOLO-ODCA network is trained and verified to obtain a trained target detection model; wherein, the target detection model includes a YOLOv7 network, a full-dimensional dynamic convolution module, a coordinate attention module, and a normalized distance module. It should be understood that in a specific embodiment, six circular targets are respectively posted on three different concrete components, and then a pair of binocular cameras KS861 with a focal length of 3.6mm are used to collect 300 artificial target images with a resolution of 2560*960 from different angles at a distance of about 5m from the target. In order to reduce the training cost, the images are uniformly cropped to images with a pixel size of 960*960, and 240 images are selected as training images, 60 as verification images, and 60 as test images. After collecting the artificial target images, they are manually annotated using the image annotation tool LabelImg to obtain data labels for subsequent network training. The construction process of the elliptical target dataset is as follows: Figure 5 shown.

[0056] Furthermore, regarding the target detection algorithm, real-time and accurate identification of elliptical targets in monitoring images has become an indispensable prerequisite for accurately estimating the pose of components. For this reason, this embodiment develops a single-stage target detection algorithm for detecting small-sized targets in complex environments based on the YOLOv7 network. This lightweight algorithm consumes less computing resources and can complete the detection of elliptical targets for real-time monitoring of the pose of hoisting components while being deployed in a binocular vision system. In the process of constructing the original YOLO-ODCA network, the improvement method is as follows:

[0057] (1) The first improvement (module 1) is to introduce Omni-Dimensional Dynamic Convolution (ODConv) through the Efficient Layer Aggregation Networks (ELAN) in YOLOv7. This convolution learns four types of attention of the convolution kernel in parallel along the four dimensions of the convolution kernel space (input channel, spatial size, output channel and number of convolution kernels), thereby improving the ability to fully extract the spatial features of the convolution kernel;

[0058] (2) The second improvement (module 2) is to introduce the coordinate attention (CA) mechanism by the last concat structure of the sampling structure in the neck network. This attention mechanism encodes position information by fusing shallow and deep feature information, capturing cross-channel information, direction perception and position-sensitive information, and improving the accuracy of positioning and identifying targets;

[0059] (3) In the target detection stage, the normalized Gaussian Wasserstein distance (NWD) is used instead of the original IoU (Intersection over Union) as the metric to solve the problems of false detection and scale sensitivity that may be caused by the IoU-based non-maximum suppression algorithm in small object detection. NWD-Loss is introduced on the basis of CIoU-Loss to provide the gradient required for network optimization in small object detection to accelerate convergence and enhance the overall model performance.

[0060] Step S12: When the monitoring result indicates that the circular target on the prefabricated hanging component is currently identified, the central pixel coordinates of the identified circular target are extracted based on a preset ellipse center detection method to obtain the corresponding target central pixel coordinates.

[0061] In this embodiment, combined with Figure 4As shown, when the monitoring result shows that the circular target on the prefabricated lifting component is currently identified, the central pixel coordinates of the identified circular target are extracted based on a preset ellipse center detection method to obtain the corresponding target center pixel coordinates, including: when the monitoring result shows that the circular target on the prefabricated lifting component is currently identified, the area where the circular target is located is cropped out from the image according to the pixel coordinates of the current detection frame to obtain the corresponding area to be detected; the area to be detected is subjected to ellipse center detection based on a preset ellipse center detection method to obtain the target center pixel coordinates corresponding to the identified circular target. That is, after the ellipse target is identified and classified, the area can be cropped out from the picture according to the pixel coordinates of the detection frame, and the cropped area is subjected to ellipse center detection to extract the pixel coordinates of the ellipse center. The detection schematic diagram is shown in FIG. Figure 6 As shown in the figure, the ellipse center detection in the entire image can be avoided, which can greatly reduce the false detection rate and improve the detection efficiency.

[0062] Step S13, based on the three-dimensional measurement principle of binocular vision and the target center pixel coordinates, the three-dimensional world coordinates of the target center point are reconstructed to obtain the corresponding three-dimensional world coordinates of the target.

[0063] In this embodiment, the three-dimensional measurement principle based on binocular vision and the target center pixel coordinates are used to reconstruct the three-dimensional world coordinates of the target center point to obtain the corresponding target three-dimensional world coordinates, including: by calibrating the binocular camera, obtaining the intrinsic parameter matrix and the extrinsic parameter matrix of the left camera and the right camera in the binocular camera, and the position conversion coefficient between the left camera and the right camera; wherein the extrinsic parameter matrix includes a rotation matrix and a displacement matrix for characterizing the coordinate conversion from the world coordinate system to the camera coordinate system; based on the three-dimensional measurement principle of binocular vision, rays are drawn using the target center pixel coordinates and the intrinsic parameter matrix, the extrinsic parameter matrix, and the position conversion coefficient, and the intersection of the rays in the space coordinate system is used as the target three-dimensional world coordinates.

[0064] That is, after obtaining the target center pixel coordinates, it is necessary to use the binocular stereo vision principle to calculate the three-dimensional world coordinates of the target point. Specifically, the conversion between the target image pixel coordinates and the world coordinates can be achieved through the imaging model of the binocular camera, and the specific expression is as follows.

[0065] (1);

[0066] In the formula, Z p is the depth of the target point; and are the pixel coordinates and world coordinates of the target point respectively; is the built-in parameter matrix of the stereo camera, where f x is the pixel length of the left camera focal length, f y is the pixel length of the right camera focal length, u0 and v0 represent the coordinates of the camera photosensitive plate center in the pixel coordinate system; and is the rotation matrix and translation matrix from the world coordinate system to the camera coordinate system, where the element a of the rotation matrix 11 、a 12 、a 13 They represent the projection of the x-axis, y-axis, and z-axis unit vectors in the world coordinate system on the x-axis of the camera coordinate system. 21 、a 22 、a 23 They represent the projection of the x, y, and z axis unit vectors in the world coordinate system on the y axis of the camera coordinate system, respectively. 31 、a 32 、a 33 They represent the projection of the x, y, and z axis unit vectors in the world coordinate system on the z axis of the camera coordinate system; t x ,t y ,t z They represent the movement distances in the directions of the three coordinate axes x, y, and z respectively, and can be read directly from the translation matrix.

[0067] Contains the rotation angles around the x, y, and z axes, which correspond to the pitch angle (rotation around the x-axis), roll angle (rotation around the y-axis), and heading angle (rotation around the z-axis), respectively.

[0068] According to the coordinate transformation relationship in the above formula, it is necessary to obtain the internal and external parameters of each binocular camera and the position conversion coefficient between the two cameras. These coefficients need to be obtained by calibrating the binocular camera. In a specific embodiment, based on the principle of Zhang Dingyou's calibration method, Matlab is used to calibrate the binocular camera. This method uses a chessboard as a reference, establishes a corresponding relationship between the vertices of the chessboard and the corresponding points on the image, and uses the known information of the chessboard (the size and number of the chessboard) to solve the internal and external parameters and distortion coefficients of the camera model. The specific steps of calibration are as follows: First, collect the calibration reference (chessboard) image pairs at different angles. For example, in order to calibrate a binocular stereo vision system with a working distance of 4 meters, a chessboard (size 30mm*30mm) made of handheld A4 paper at multiple points within 4 meters is photographed by a binocular camera, and a total of 59 pairs of chessboard images at different distances and angles are collected. Then, the collected chessboard image pairs are input into the Matlab binocular calibration program to extract the corner points of the chessboard. The calibration results can be viewed through the reprojection error histogram, such as Figure 7As shown in the public reprojection error diagram, the reprojection error value is generally controlled within 0.1mm. Finally, the distortion coefficient and internal and external parameters of the binocular camera are analyzed by the maximum likelihood estimation method.

[0069] The spatial coordinate calculation method used in this embodiment is based on the triangulation principle of binocular vision. Figure 8 As shown. During the camera imaging process, spatial information is lost and only two-dimensional information on the imaging plane is retained. In order to calculate the spatial coordinates, a stereoscopic vision system is formed by combining two or more cameras. This paper uses left and right cameras to construct a binocular stereoscopic vision system. By calculating the pixel coordinates of the same monitoring point in the left and right cameras, the optical centers of the left and right cameras and the pixel coordinates of the monitoring points under the cameras are drawn as rays, and the intersection of these rays in the spatial coordinate system is the required spatial coordinate point. Camera imaging is a projection process, which has been mentioned in equation (1). In order to derive the triangulation formula, equation (1) is expanded to obtain a more specific camera imaging model, and the specific expression is shown below.

[0070] (2);

[0071] Where u, v are the pixel coordinates of the monitored target; X W , Y W , Z W They respectively represent the coordinates of the monitored target on the X, Y, and Z axes of the left camera coordinate system; the product of the camera's intrinsic parameter matrix and extrinsic parameter matrix can be recorded as the new matrix M, then the simplified imaging models of the left camera and the right camera are shown below.

[0072] (3);

[0073] (4);

[0074] In the formula, (u1,v1) (that is, Figure 8 Point p in l (u l ,v l )) and (u2,v2) (that is, Figure 8 Point p in r (u r ,v r )) represent the pixel coordinates of the monitored target in the left camera and the right camera respectively, and S1 and S2 are two non-zero scalars. Eliminating the two scalars S1 and S2 from the equation, the linear equation can be transformed into:

[0075] (5); (6);

[0076] Since a linear equation represents a plane in space and two composite linear equations represent an intersection of two planes, equations (5) and (6) refer to two lines in three-dimensional (3D) space that intersect at point P(X W , Y W , Z W ), which is the target 3D world coordinate. Obviously, it is difficult for two spatial lines to intersect exactly at a spatial point. Therefore, it is necessary to seek a least squares solution to find a point that minimizes the sum of the distances from the point to the two straight lines.

[0077] Step S14, by using the target three-dimensional world coordinates to construct a component coordinate system and a reference coordinate system representing the current state and the design state respectively, obtain the current posture information of the prefabricated lifting component relative to the design state, so as to guide the installation of the component based on the posture information.

[0078] In this embodiment, since the control point is usually located at a position that is difficult to directly observe, such as the cross-sectional centroid of the bottom section, a circular target is pre-installed on the visible surface of the prefabricated lifting component. In order to estimate the posture of the prefabricated lifting component and provide guidance for accurate installation, the process can be divided into the following two steps:

[0079] (1) Using the spatial coordinates of the target point in the L-CCS (Left Camera Coordinate System) to construct the component coordinate system (CCS), i.e., the design coordinate system and the reference coordinate system (RCS), then converting the relative pose measurement between the prefabricated hoisting component and the existing structure (i.e., the existing structure in the figure) into a process of determining the relative relationship between the two coordinate systems, i.e., determining the corresponding transformation matrices R1 and T1;

[0080] (2) According to the structural design information, the key points that need to be controlled are selected, the coordinates of the key points in the CCS are calculated, and then converted into RCS coordinate values ​​based on R1 and T1, and deviation calculation is performed to output the vertical and horizontal deviation information of the key points as posture information, and the installation is directly guided based on the vertical and horizontal deviation information.

[0081] That is, if Fig. 9As shown in the figure, the three-dimensional coordinates of the pre-installed target are measured, and the reference coordinate system RCS (o1-xyz) is established on the existing structure, and CCS (o1-xyz) is established on the prefabricated hoisting component. The center of mass E and the four corners A, B, C, and D are selected as the key control points on the bottom surface, and E1, A1, B1, C1, and D1 are their corresponding design states on the existing structure. First, the spatial coordinates of the key points on the prefabricated hoisting component are calculated in CCS according to the structural design information, and then converted into RCS for deviation calculation.

[0082] In summary, this embodiment proposes a method for monitoring the posture of a hoisting component using an artificial target and binocular vision, which can continuously collect data and output the posture of the hoisting component relative to the design state online. An improved target detection algorithm based on deep learning is also proposed, which can realize real-time and accurate recognition and classification of small-sized elliptical targets pre-installed on hoisting components in complex hoisting scenarios. In addition, an improved high-quality ellipse detection algorithm is also proposed to facilitate the extraction of pixel coordinates of the center of the elliptical target. This avoids ellipse center detection in the entire image, which can greatly reduce the false detection rate and improve detection efficiency.

[0083] It is important to understand that, since the binocular vision system allows continuous image acquisition and online data transmission throughout the entire assembly process, the crane operator can precisely adjust the precast concrete components in real time according to the estimated spatial pose relative to the design state.

[0084] It can be seen that in the embodiment of the present application, in the current lifting bridge component assembly scene, a binocular vision system in a tripod at a preset distance from the prefabricated lifting component is used to monitor the lifting site where a circular target has been preinstalled on the prefabricated lifting component and the target monitoring position in real time, and target recognition is performed on the monitoring image to obtain corresponding monitoring results; wherein the binocular vision system is a system equipped with a binocular camera and a pre-trained target detection model; when the monitoring result shows that the circular target on the prefabricated lifting component is currently identified, the center pixel coordinates of the identified circular target are extracted based on a preset ellipse center detection method to obtain the corresponding target center pixel coordinates; the three-dimensional world coordinates of the target center point are reconstructed based on the three-dimensional measurement principle of binocular vision and the target center pixel coordinates to obtain the corresponding target three-dimensional world coordinates; by using the target three-dimensional world coordinates to construct a component coordinate system and a reference coordinate system representing the current state and the design state respectively, the current posture information of the prefabricated lifting component relative to the design state is obtained, so as to guide the installation of components based on the posture information. That is, in this application, a binocular vision system equipped with a binocular camera and a pre-trained target detection model monitors the hoisting site in real time, and when the circular target pre-installed on the prefabricated hoisting component is detected, the target center pixel coordinates are extracted based on the preset ellipse center detection method, and the corresponding target three-dimensional world coordinates are obtained based on the three-dimensional measurement principle of binocular vision, so as to obtain the current posture information of the prefabricated hoisting component relative to the design state, so as to guide the installation of the component based on the posture information. In this way, the posture information of the prefabricated hoisting component can be effectively monitored in real time and accurately to guide the installation of the component while improving efficiency, and avoid unnecessary waste of human resources.

[0085] See also Fig.10 As shown, an embodiment of the present invention discloses a method for intelligently monitoring the posture of a precast concrete component, comprising:

[0086] Step S21: In the current lifting bridge component assembly scene, a binocular vision system in a tripod at a preset distance from the prefabricated lifting component is used to perform real-time monitoring of the lifting site where circular targets have been preinstalled on the prefabricated lifting components and the target monitoring position, and target recognition is performed on the monitoring image to obtain corresponding monitoring results; wherein the binocular vision system is a system equipped with a binocular camera and a pre-trained target detection model.

[0087] Step S22: When the monitoring result indicates that the circular target on the prefabricated hanging component is currently identified, the central pixel coordinates of the identified circular target are extracted based on a preset ellipse center detection method to obtain corresponding target central pixel coordinates.

[0088] Step S23, based on the three-dimensional measurement principle of binocular vision and the target center pixel coordinates, the three-dimensional world coordinates of the target center point are reconstructed to obtain the corresponding three-dimensional world coordinates of the target.

[0089] Step S24, using the target three-dimensional world coordinates to construct a component coordinate system representing the current state on the prefabricated lifting component, and to construct a reference coordinate system representing the design state on the existing structure; wherein the existing structure is a structure that has been installed in the current lifting bridge component assembly scene and is designed to install the prefabricated lifting component.

[0090] Step S25: Select a corresponding target control point according to the structural design information, and calculate the target coordinate information of the target control point in the component coordinate system.

[0091] Step S26: converting the target coordinate information in the reference coordinate system and performing deviation calculation to obtain the current posture information of the prefabricated lifting component relative to the design state, and outputting the posture information online; wherein the posture information includes corresponding vertical deviation information and horizontal deviation information, so as to guide the installation of the component based on the posture information.

[0092] Among them, the specific process of the above steps S21 to S26 can refer to the corresponding content disclosed in the above embodiments, and will not be repeated here.

[0093] It can be seen that in the embodiment of the present application, after obtaining the target three-dimensional world coordinates, the relative relationship between the two coordinate systems is determined by converting the relative posture measurement values ​​between the prefabricated hoisting component and the existing structure. Then, by converting all the measurement values ​​into a common value, that is, into RCS coordinate values, the vertical and horizontal deviations of the key points can be automatically and continuously output, thereby directly guiding the installation.

[0094] In the current lifting bridge component assembly scenario, a binocular vision system in a tripod at a preset distance from the prefabricated lifting component is used to monitor the lifting site in real time, where a circular target has been preinstalled on the prefabricated lifting component and the target monitoring position, and target recognition is performed on the monitoring image to obtain corresponding monitoring results; wherein the binocular vision system is a system equipped with a binocular camera and a pre-trained target detection model; when the monitoring result shows that the circular target on the prefabricated lifting component is currently identified, the central pixel coordinates of the identified circular target are extracted based on a preset ellipse center detection method to obtain the corresponding target center pixel coordinates; the three-dimensional world coordinates of the target center point are reconstructed based on the three-dimensional measurement principle of binocular vision and the target center pixel coordinates to obtain the corresponding target three-dimensional world coordinates; by using the target three-dimensional world coordinates to construct a component coordinate system and a reference coordinate system representing the current state and the design state respectively, the current posture information of the prefabricated lifting component relative to the design state is obtained, so as to guide the installation of the component based on the posture information. That is, in this application, a binocular vision system equipped with a binocular camera and a pre-trained target detection model monitors the hoisting site in real time, and when the circular target pre-installed on the prefabricated hoisting component is detected, the target center pixel coordinates are extracted based on the preset ellipse center detection method, and the corresponding target three-dimensional world coordinates are obtained based on the three-dimensional measurement principle of binocular vision, so as to obtain the current posture information of the prefabricated hoisting component relative to the design state, so as to guide the installation of the component based on the posture information. In this way, the posture information of the prefabricated hoisting component can be effectively monitored in real time and accurately to guide the installation of the component while improving efficiency, and avoid unnecessary waste of human resources.

[0095] See also Fig.11 As shown, the embodiment of the present application also discloses a precast concrete component posture intelligent monitoring system, including:

[0096] The real-time monitoring module 11 is used to perform real-time monitoring of the hoisting site where circular targets have been pre-installed at the prefabricated hoisting components and the target monitoring position through a binocular vision system in a tripod at a preset distance from the prefabricated hoisting components in the current hoisting bridge component assembly scene, and perform target recognition on the monitoring image to obtain corresponding monitoring results; wherein the binocular vision system is a system equipped with a binocular camera and a pre-trained target detection model;

[0097] A center coordinate acquisition module 12 is used to extract the center pixel coordinates of the identified circular target based on a preset ellipse center detection method to obtain the corresponding target center pixel coordinates when the monitoring result indicates that the circular target on the prefabricated hoisting component is currently identified;

[0098] A three-dimensional coordinate acquisition module 13 is used to reconstruct the three-dimensional world coordinates of the target center point based on the three-dimensional measurement principle of binocular vision and the target center pixel coordinates to obtain the corresponding target three-dimensional world coordinates;

[0099] The posture information acquisition module 14 is used to construct a component coordinate system and a reference coordinate system representing the current state and the design state respectively by using the target three-dimensional world coordinates, and obtain the current posture information of the prefabricated lifting component relative to the design state, so as to guide the installation of the component based on the posture information.

[0100] Among them, for more specific working processes of the above-mentioned modules, please refer to the corresponding contents disclosed in the aforementioned embodiments, which will not be repeated here.

[0101] It can be seen that in the current lifting bridge component assembly scenario, the present application uses a binocular vision system in a tripod at a preset distance from the prefabricated lifting component to conduct real-time monitoring of the lifting site where a circular target has been pre-installed on the prefabricated lifting component and the target monitoring position, and performs target recognition on the monitoring image to obtain corresponding monitoring results; wherein the binocular vision system is a system equipped with a binocular camera and a pre-trained target detection model; when the monitoring result shows that the circular target on the prefabricated lifting component is currently identified, the center pixel coordinates of the identified circular target are extracted based on a preset ellipse center detection method to obtain the corresponding target center pixel coordinates; the three-dimensional world coordinates of the target center point are reconstructed based on the three-dimensional measurement principle of binocular vision and the target center pixel coordinates to obtain the corresponding target three-dimensional world coordinates; by using the target three-dimensional world coordinates to construct a component coordinate system and a reference coordinate system representing the current state and the design state respectively, the current posture information of the prefabricated lifting component relative to the design state is obtained, so as to guide the installation of components based on the posture information. That is, in this application, a binocular vision system equipped with a binocular camera and a pre-trained target detection model monitors the hoisting site in real time, and when the circular target pre-installed on the prefabricated hoisting component is detected, the target center pixel coordinates are extracted based on the preset ellipse center detection method, and the corresponding target three-dimensional world coordinates are obtained based on the three-dimensional measurement principle of binocular vision, so as to obtain the current posture information of the prefabricated hoisting component relative to the design state, so as to guide the installation of the component based on the posture information. In this way, the posture information of the prefabricated hoisting component can be effectively monitored in real time and accurately to guide the installation of the component while improving efficiency, and avoid unnecessary waste of human resources.

[0102] In some specific embodiments, the precast concrete component position intelligent monitoring system may further include:

[0103] The target preinstallation unit is used to preinstall preconfigured circular targets at the monitorable positions of the prefabricated hanging components and the designed reference positions by using a total station.

[0104] In some specific embodiments, the precast concrete component position intelligent monitoring system may further include:

[0105] A dataset construction unit, used to construct a dataset containing multiple images;

[0106] A data set division unit, used for dividing the data set after cropping each of the images in the data set to a uniform size, so as to obtain a corresponding training data set and a verification data set;

[0107] The model training unit is used to train and verify the constructed original YOLO-ODCA network based on the training data set and the verification data set, so as to obtain a trained target detection model; wherein the target detection model includes a YOLOv7 network, a full-dimensional dynamic convolution module, a coordinate attention module, and a normalized distance module.

[0108] In some specific embodiments, the precast concrete component position intelligent monitoring system may further include:

[0109] The original network construction unit is used to introduce full-dimensional dynamic convolution in the efficient layer aggregation network in YOLOv7, introduce the coordinate attention mechanism in the last concat structure of the sampling structure on the neck network, use the normalized Gaussian Wasserstein distance instead of the original IoU as a metric, and introduce NWD-Loss on the basis of CIoU-Loss to obtain the original YOLO-ODCA network in the process of constructing the original YOLO-ODCA network.

[0110] In some specific embodiments, the center coordinate acquisition module 12 may specifically include:

[0111] A unit for acquiring an area to be detected, configured to, when the monitoring result indicates that the circular target on the prefabricated hanging component is currently identified, cut out the area where the circular target is located from the image according to the pixel coordinates of the current detection frame to obtain a corresponding area to be detected;

[0112] The target center pixel coordinate acquisition unit is used to perform ellipse center detection on the area to be detected based on a preset ellipse center detection method to obtain the target center pixel coordinates corresponding to the identified circular target.

[0113] In some specific embodiments, the three-dimensional coordinate acquisition module 13 may specifically include:

[0114] An internal and external parameter matrix acquisition unit is used to obtain the intrinsic parameter matrix and the extrinsic parameter matrix of each of the left camera and the right camera in the binocular camera, and the position conversion coefficient between the left camera and the right camera by calibrating the binocular camera; wherein the extrinsic parameter matrix includes a rotation matrix and a displacement matrix for representing the coordinate conversion from the world coordinate system to the camera coordinate system;

[0115] The target three-dimensional world coordinate acquisition unit is used for three-dimensional measurement principle based on binocular vision, and uses the target center pixel coordinates and the intrinsic parameter matrix, the extrinsic parameter matrix, and the position conversion coefficient to draw rays, and uses the intersection of the rays in the space coordinate system as the target three-dimensional world coordinates.

[0116] In some specific embodiments, the posture information acquisition module 14 may specifically include:

[0117] A coordinate system construction unit, used to construct a component coordinate system representing the current state on the prefabricated hoisting component using the target three-dimensional world coordinates, and to construct a reference coordinate system representing the design state on the existing structure; wherein the existing structure is a structure that has been installed in the current hoisting bridge component assembly scene and is designed to install the prefabricated hoisting component;

[0118] A target coordinate acquisition unit, used to select a corresponding target control point according to the structural design information, and calculate the target coordinate information of the target control point in the component coordinate system;

[0119] The posture information acquisition unit is used to convert the target coordinate information in the reference coordinate system, perform deviation calculation, so as to obtain the current posture information of the prefabricated lifting component relative to the design state, and output the posture information online; wherein the posture information includes corresponding vertical deviation information and horizontal deviation information.

[0120] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0121] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0122] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0123] The technical solution provided by the present application is introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for general technicians in this field, according to the idea of ​​the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for intelligent monitoring of the position and posture of precast concrete components, characterized in that: include: In the current hoisting bridge component assembly scenario, a binocular vision system in a tripod at a preset distance from the prefabricated hoisting component is used to monitor the hoisting site where a circular target has been preinstalled at the prefabricated hoisting component and the target monitoring position in real time, and target recognition is performed on the monitoring image to obtain corresponding monitoring results; wherein the binocular vision system is a system equipped with a binocular camera and a pre-trained target detection model; When the monitoring result indicates that the circular target on the prefabricated hoisting component is currently identified, the central pixel coordinates of the identified circular target are extracted based on a preset ellipse center detection method to obtain corresponding target central pixel coordinates; Reconstructing the three-dimensional world coordinates of the target center point based on the three-dimensional measurement principle of binocular vision and the target center pixel coordinates to obtain the corresponding target three-dimensional world coordinates; By using the target three-dimensional world coordinates to construct a component coordinate system and a reference coordinate system representing a current state and a design state respectively, the current posture information of the prefabricated hoisting component relative to the design state is obtained, so as to guide the installation of the component based on the posture information; Wherein, before the binocular vision system in the tripod at a preset distance from the prefabricated hoisting component is used to conduct real-time monitoring of the hoisting site where the circular target has been preinstalled on the prefabricated hoisting component and the target monitoring position, it also includes: Build a dataset containing multiple images; After cropping each of the images in the data set to a uniform size, dividing the data set to obtain a corresponding training data set and a verification data set; Based on the training data set and the verification data set, the constructed original YOLO-ODCA network is trained and verified respectively to obtain a trained target detection model; wherein the target detection model includes a YOLOv7 network, a full-dimensional dynamic convolution module, a coordinate attention module, and a normalized distance module; Before respectively training and verifying the constructed original YOLO-ODCA network based on the training data set and the verification data set, the method further includes: In the process of constructing the original YOLO-ODCA network, the full-dimensional dynamic convolution is introduced into the efficient layer aggregation network in YOLOv7, the coordinate attention mechanism is introduced into the last concat structure of the sampling structure on the neck network, the normalized Gaussian Wasserstein distance is used instead of the original IoU as the metric, and NWD-Loss is introduced on the basis of CIoU-Loss to obtain the original YOLO-ODCA network.

2. The method for intelligent monitoring of the position and posture of precast concrete components according to claim 1 is characterized in that: Also includes: A total station is used to pre-install pre-configured circular targets at the monitorable locations of the prefabricated hanging components and the designed reference positions.

3. The method for intelligent monitoring of the position and posture of precast concrete components according to claim 1, characterized in that: When the monitoring result indicates that the circular target on the prefabricated hoisting component is currently identified, extracting the central pixel coordinates of the identified circular target based on a preset ellipse center detection method to obtain the corresponding target central pixel coordinates, including: When the monitoring result indicates that the circular target on the prefabricated hanging component is currently identified, the area where the circular target is located is cropped out from the image according to the pixel coordinates of the current detection frame to obtain a corresponding area to be detected; Based on a preset ellipse center detection method, an ellipse center detection is performed on the area to be detected to obtain target center pixel coordinates corresponding to the identified circular target.

4. The method for intelligent monitoring of the position and posture of precast concrete components according to any one of claims 1 to 3, characterized in that: The three-dimensional measurement principle based on binocular vision and the target center pixel coordinates are used to reconstruct the three-dimensional world coordinates of the target center point to obtain the corresponding target three-dimensional world coordinates, including: By calibrating the binocular camera, an intrinsic parameter matrix and an extrinsic parameter matrix of each of a left camera and a right camera in the binocular camera, and a position conversion coefficient between the left camera and the right camera are obtained; wherein the extrinsic parameter matrix includes a rotation matrix and a displacement matrix for representing the conversion of coordinates in a world coordinate system to a camera coordinate system; Based on the three-dimensional measurement principle of binocular vision, rays are drawn using the target center pixel coordinates, the intrinsic parameter matrix, the extrinsic parameter matrix, and the position conversion coefficient, and the intersection of the rays in the space coordinate system is used as the target three-dimensional world coordinates.

5. The method for intelligent monitoring of the position and posture of precast concrete components according to claim 4 is characterized in that: The method of constructing a component coordinate system and a reference coordinate system respectively representing a current state and a design state by using the target three-dimensional world coordinates to obtain the current position information of the prefabricated hoisting component relative to the design state includes: A component coordinate system representing the current state is constructed on the prefabricated hoisting component using the target three-dimensional world coordinates, and a reference coordinate system representing the design state is constructed on the existing structure; wherein the existing structure is a structure that has been installed in the current hoisting bridge component assembly scene and is designed to install the prefabricated hoisting component; Selecting a corresponding target control point according to the structural design information, and calculating target coordinate information of the target control point in the component coordinate system; The target coordinate information is converted in the reference coordinate system and deviation is calculated to obtain the current posture information of the prefabricated hoisting component relative to the design state, and the posture information is output online; wherein the posture information includes corresponding vertical deviation information and horizontal deviation information.

6. An intelligent monitoring system for the position and posture of precast concrete components, characterized in that: include: A real-time monitoring module is used to perform real-time monitoring of the hoisting site where circular targets have been pre-installed at the prefabricated hoisting components and target monitoring positions through a binocular vision system in a tripod at a preset distance from the prefabricated hoisting components in the current hoisting bridge component assembly scene, and to perform target recognition on the monitoring image to obtain corresponding monitoring results; wherein the binocular vision system is a system equipped with a binocular camera and a pre-trained target detection model; A center coordinate acquisition module, for extracting the center pixel coordinates of the identified circular target based on a preset ellipse center detection method to obtain the corresponding target center pixel coordinates when the monitoring result indicates that the circular target on the prefabricated hoisting component is currently identified; A three-dimensional coordinate acquisition module is used to reconstruct the three-dimensional world coordinates of the target center point based on the three-dimensional measurement principle of binocular vision and the target center pixel coordinates to obtain the corresponding target three-dimensional world coordinates; A posture information acquisition module is used to construct a component coordinate system and a reference coordinate system representing a current state and a design state respectively by using the target three-dimensional world coordinates, so as to obtain the current posture information of the prefabricated hoisting component relative to the design state, so as to guide the installation of the component based on the posture information; Among them, it also includes: A dataset construction unit, used to construct a dataset containing multiple images; A data set division unit, used for dividing the data set after cropping each of the images in the data set to a uniform size, so as to obtain a corresponding training data set and a verification data set; A model training unit, used to train and verify the constructed original YOLO-ODCA network based on the training data set and the verification data set, respectively, to obtain a trained target detection model; wherein the target detection model includes a YOLOv7 network, a full-dimensional dynamic convolution module, a coordinate attention module, and a normalized distance module; Also includes: The original network construction unit is used to introduce full-dimensional dynamic convolution in the efficient layer aggregation network in YOLOv7, introduce the coordinate attention mechanism in the last concat structure of the sampling structure on the neck network, use the normalized Gaussian Wasserstein distance instead of the original IoU as a metric, and introduce NWD-Loss on the basis of CIoU-Loss to obtain the original YOLO-ODCA network in the process of constructing the original YOLO-ODCA network.