Machine vision-combined ALC (autoclaved lightweight concrete) internal partition wall board assembly control method and system

Abstract

The invention provides an ALC internal partition wall board splicing control method and system combined with machine vision, and relates to the technical field of visual inspection, and the method comprises the steps: obtaining a wallboard construction robot, and collecting current wallboard appearance image data and wallboard laser point cloud data through a binocular camera and a laser radar; performing defect detection and visual positioning on the current wallboard appearance image data and the wallboard laser point cloud data according to a wallboard assembly application standard; three-dimensional point cloud data of a construction site are collected, and splicing path planning is carried out on space positions of available wallboards; the current wallboard appearance image data is controlled and analyzed based on a force sensor, and a wallboard construction robot is controlled to conduct ALC inner partition wallboard splicing control based on the wallboard splicing control parameters and the target wallboard splicing path. The technical problem that in the prior art, due to the fact that manual splicing lacks accurate detection and positioning, the wallboard splicing precision is insufficient is solved, and the splicing precision is improved by integrating machine vision.

Description

Technical Field

[0001] This application relates to the field of visual inspection technology, and in particular to an ALC interior partition panel assembly control method and system that combines machine vision. Background Technology

[0002] In the traditional construction process of ALC interior partition wall panels, the assembly quality highly depends on the individual experience and skill level of the construction workers. Due to the lack of objective and unified quantitative control standards, significant individual differences exist in key aspects such as panel positioning, alignment, splicing, and compaction. Manual assembly makes it difficult to ensure the accuracy of the wall panels, especially in terms of panel position, verticality, and horizontality, which can easily lead to instability in the overall structure due to the accumulation of errors. Furthermore, the lack of defect detection for ALC interior partition wall panels, the difficulty in controlling core physical parameters during assembly, and the lack of high-precision environmental perception and automated path planning result in insufficient wall panel assembly accuracy, leading to unstable assembly quality of ALC interior partition wall panels.

[0003] In summary, the existing technology has a technical problem: due to the lack of precise detection and positioning during manual assembly, the wall panel assembly accuracy is insufficient, which further affects the assembly quality of the ALC interior partition wall panels. Summary of the Invention

[0004] The purpose of this application is to provide an ALC interior partition wall panel assembly control method and system that combines machine vision, in order to solve the technical problem in the prior art that the lack of accurate detection and positioning in manual assembly leads to insufficient wall panel assembly accuracy, which further affects the assembly quality of ALC interior partition wall panels.

[0005] In view of the above problems, this application provides an ALC internal partition wall panel assembly control method and system that combines machine vision.

[0006] Firstly, this application provides an ALC internal partition wall panel assembly control method incorporating machine vision. This method is implemented through an ALC internal partition wall panel assembly control system incorporating machine vision. The method includes: acquiring a wall panel construction robot, which integrates a binocular camera, a lidar, and a force sensor; acquiring current wall panel appearance image data and wall panel laser point cloud data through the binocular camera and lidar; and assembling the current wall panel according to wall panel assembly application standards. Defect detection and visual positioning are performed using appearance image data and wall panel laser point cloud data to determine the available wall panel spatial location; three-dimensional point cloud data of the construction site is collected, and assembly path planning is performed on the available wall panel spatial location based on the three-dimensional point cloud data of the construction site to determine the target wall panel assembly path; the wall panel construction robot is controlled and analyzed based on the force sensor of the current wall panel appearance image data to determine the wall panel assembly control parameters, and the wall panel construction robot is controlled to perform ALC internal partition wall panel assembly control based on the wall panel assembly control parameters and the target wall panel assembly path.

[0007] Optionally, standard wall panel appearance image data is determined according to the wall panel assembly application standard; defect comparison and detection are performed on the current wall panel appearance image data according to the standard wall panel appearance image data to obtain the current wall panel defect detection result; when the current wall panel defect detection result reaches the preset unqualified defect threshold, the current wall panel is marked as unqualified and automatically skipped; if the current wall panel defect detection result does not reach the preset unqualified defect threshold, visual positioning is performed based on the wall panel laser point cloud data to determine the available wall panel spatial location.

[0008] Optionally, multi-dimensional feature extraction is performed on the standard wall panel appearance image data and the current wall panel appearance image data respectively to obtain the standard wall panel appearance image feature set and the current wall panel appearance image feature set; loss detection comparison is performed on the current wall panel appearance image feature set according to the standard wall panel appearance image feature set to obtain the current wall panel loss feature set; a wall panel defect assessment system is built, and the defect degree of the current wall panel loss feature set is assessed using the wall panel defect assessment system to obtain the current wall panel defect detection result.

[0009] Optionally, the laser point cloud data of the wall panel is filtered and denoised, and key feature points are extracted to obtain a set of key feature points of the wall panel; the key feature point set of the wall panel is fitted with edge contours according to the standard wall panel appearance image data to generate wall panel edge contour information; the wall panel edge contour information is used to estimate the pose of the laser point cloud data of the wall panel to determine the wall panel placement pose information, and the wall panel placement pose information is mapped and transformed to the world coordinate system to determine the spatial position of the available wall panel.

[0010] Optionally, based on the three-dimensional point cloud data of the construction site, filtering and denoising are performed to generate a spatial model of the wall panel construction site; the wall panel assembly area is marked on the spatial model of the wall panel construction site to determine the target wall panel assembly space location; based on the spatial model of the wall panel construction site, the available wall panel space location is used as the starting point and the target wall panel assembly space location is used as the ending point to plan the assembly path and obtain the target wall panel assembly path.

[0011] Optionally, the motion constraint parameters of the wall panel construction robot are obtained, and obstacle identification is performed on the wall panel construction site space model based on the motion constraint parameters to generate an available wall panel construction space model; a wall panel assembly path cost function is constructed according to the internal partition wall panel assembly target; the available wall panel space location is taken as the starting point, and the target wall panel assembly space location is taken as the ending point, and path planning is performed on the wall panel assembly path cost function based on the wall panel construction site space model to obtain the target wall panel assembly path.

[0012] Optionally, the standard wall panel weight information of the current wall panel appearance image data is obtained, and the standard wall panel weight information is corrected for loss based on the current wall panel loss feature set to obtain the current wall panel weight information; the robot assembly contact point is determined according to the current wall panel appearance image data; the clamping force constraint parameters of the wall panel construction robot are obtained, and the robot assembly contact point is controlled and analyzed based on the force sensor using the motion constraint parameters, the clamping force constraint parameters, and the current wall panel weight information to obtain the wall panel assembly control parameters.

[0013] Optionally, based on the motion constraint parameters, the clamping force constraint parameters, and the current wall panel weight information, the robot assembly contact point is analyzed for assembly control to obtain initial assembly control parameters; the wall panel construction robot executes wall panel assembly control based on the initial assembly control parameters, and simultaneously collects clamping force sensing data during the assembly process through the force sensor; the initial assembly control parameters are adjusted and optimized based on the clamping force sensing data to determine the wall panel assembly control parameters.

[0014] Optionally, steady-state over-limit detection and feedback optimization analysis are performed on the clamping force sensing data to obtain the force feedback optimization direction; the initial assembly control parameters are iteratively adjusted and optimized based on the force feedback optimization direction to determine the wall panel assembly control parameters.

[0015] Secondly, this application also provides an ALC interior partition wall panel assembly control system incorporating machine vision, used to execute the ALC interior partition wall panel assembly control method incorporating machine vision as described in the first aspect, wherein the ALC interior partition wall panel assembly control system incorporating machine vision includes: a robot determination module, used to acquire a wall panel construction robot, the wall panel construction robot integrating a binocular camera, a lidar, and a force sensor, and acquiring current wall panel appearance image data and wall panel laser point cloud data through the binocular camera and lidar; and a vision inspection module, used to inspect the current wall panel appearance image data according to wall panel assembly application standards. The system performs defect detection and visual positioning using laser point cloud data of the wall panels to determine the available spatial location of the wall panels; the path planning module is used to collect three-dimensional point cloud data of the construction site, and plan the assembly path of the available wall panels based on the three-dimensional point cloud data of the construction site to determine the assembly path of the target wall panels; the assembly control module is used to control and analyze the current wall panel appearance image data based on the force sensor through the wall panel construction robot, determine the wall panel assembly control parameters, and control the wall panel construction robot to perform ALC internal partition wall panel assembly control based on the wall panel assembly control parameters and the target wall panel assembly path.

[0016] One or more technical solutions provided in this application have at least the following beneficial effects: A wall panel construction robot is acquired, which integrates a binocular camera, a lidar, and a force sensor. The binocular camera and lidar acquire current wall panel appearance image data and wall panel laser point cloud data. Defect detection and visual positioning are performed on the current wall panel appearance image data and wall panel laser point cloud data according to wall panel assembly application standards to determine the available wall panel spatial location. Three-dimensional point cloud data of the construction site is acquired, and assembly path planning is performed on the available wall panel spatial location based on the three-dimensional point cloud data of the construction site to determine the target wall panel assembly path. The wall panel construction robot performs control analysis on the current wall panel appearance image data based on the force sensor to determine wall panel assembly control parameters, and controls the wall panel construction robot to perform ALC internal partition wall panel assembly control based on the wall panel assembly control parameters and the target wall panel assembly path. In other words, data acquired using binocular cameras and lidar is used to determine whether there are quality problems with the wall panels through defect detection and visual positioning, and to determine whether the wall panels can be used for assembly. The three-dimensional point cloud data of the construction site is used to plan the assembly path of the wall panels, ensuring that the spatial layout during the assembly process meets the design requirements. Based on the feedback from force sensors and visual data, assembly control parameters are formulated to control the wall panel construction robot to accurately assemble the wall panels, avoid deviations or damage, reduce assembly errors, and improve the assembly quality of the ALC interior partition wall panels.

[0017] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the ALC internal partition panel assembly control method that incorporates machine vision, as described in this application.

[0020] Figure 2 This is a schematic diagram of the ALC internal partition wall panel assembly control system that incorporates machine vision, as described in this application.

[0021] Explanation of reference numerals in the attached diagram: Robot determination module 11, vision inspection module 12, path planning module 13, assembly control module 14. Detailed Implementation

[0022] This application provides a machine vision-based ALC (Automatic Classified Concrete) interior partition wall panel assembly control method and system, solving the technical problem in existing technologies where insufficient precision in wall panel assembly due to the lack of accurate detection and positioning during manual assembly leads to insufficient assembly accuracy, further affecting the assembly quality of ALC interior partition wall panels. Using data acquired by binocular cameras and LiDAR, defect detection and visual positioning are used to determine whether wall panels have quality problems and whether they can be used for assembly. Three-dimensional point cloud data from the construction site is used to plan the wall panel assembly path, ensuring that the spatial layout during assembly meets design requirements. Based on force sensor feedback and visual data, assembly control parameters are formulated to control the wall panel construction robot to precisely assemble the wall panels, avoiding deviations or damage, reducing assembly errors, and improving the assembly quality of ALC interior partition wall panels.

[0023] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0024] Example 1, please refer to the appendix. Figure 1 This application provides a machine vision-based ALC internal partition panel assembly control method, wherein the machine vision-based ALC internal partition panel assembly control method is executed by a machine vision-based ALC internal partition panel assembly control system, and the machine vision-based ALC internal partition panel assembly control method specifically includes the following steps: A wall panel construction robot is acquired. The wall panel construction robot integrates a binocular camera, a lidar, and a force sensor. The binocular camera and lidar are used to collect current wall panel appearance image data and wall panel laser point cloud data.

[0025] Specifically, ALC interior partition wall panels, also known as autoclaved aerated concrete (AAC) panels, are a new type of building material, serving as prefabricated interior partition wall panels. ALC interior partition wall panels are prefabricated with precise dimensions, using a 600mm wide modular design. The height can be customized according to the floor height, facilitating design and construction. Unlike traditional brick or block walls constructed with mortar, ALC interior partition wall panels are transported to the construction site as finished products and then assembled using specialized installation equipment and connecting materials to form complete interior partition walls.

[0026] A wall panel construction robot is a specialized robot designed for the automated installation of ALC (Alternating Current Concrete) interior partition walls. It typically consists of a movable base, a multi-degree-of-freedom robotic arm, a drive module, a sensor system, and a control system. It can replace manual labor in tasks such as panel handling, positioning, and assembly. The wall panel construction robot integrates a binocular camera, LiDAR, and force sensors. The binocular camera, a vision sensor composed of two cameras, simulates human eyes. By calculating the pixel differences (parallax) of the same object in the images from the two cameras, it obtains the object's depth information, forming an RGB-D image (color + depth map) used to acquire the object's texture, color, and two-dimensional geometric information. The LiDAR uses a laser beam to scan the target and acquires spatial three-dimensional point cloud data by measuring the laser's round-trip time. This accurately describes the construction site environment and the spatial shape of the ALC interior partition walls, generating precise three-dimensional point cloud data. The force sensor detects the forces and torques acting on the wall panel construction robot during operation, acting as its tactile sense, enabling it to perceive contact forces with the environment and achieve compliant and precise control.

[0027] The binocular camera and LiDAR sensor driving the wall panel construction robot are aimed at the ALC interior partition wall panel to be grasped. The binocular camera simultaneously captures high-definition color images of the ALC interior partition wall panel from two slightly different angles, calculates a disparity map using a stereo vision algorithm, and then generates 3D visual perception data containing depth information. Simultaneously, the LiDAR performs a high-speed laser scan of the ALC interior partition wall panel surface in the same area, directly acquiring massive amounts of precise 3D point data using the time-of-flight principle, forming a high-precision point cloud model. The current wall panel appearance image data is a high-resolution RGB color image of the ALC interior partition wall panel to be installed, acquired by the binocular camera, containing information on the texture, color, edges, and visible defects such as cracks, chipped edges, and oil stains on the surface of the ALC interior partition wall panel. The wall panel laser point cloud data is a set of data acquired by the LiDAR after scanning the ALC interior partition wall panel to be installed; each point contains 3D coordinates (X, Y, Z) and laser reflection intensity information. Tens of thousands of points together constitute the three-dimensional geometry of the ALC interior partition wall surface, accurately describing its shape, size, flatness, and orientation.

[0028] In one example, in an office building project with a floor height of 2.95 meters, a wall panel construction robot moved to the stacking area and targeted a standard ALC (Alternating Current Concrete) interior partition wall panel with dimensions of 600mm × 2950mm × 100mm. A binocular camera with a resolution of 1280*720 captured an image at a distance of approximately 1.2 meters from the surface of the ALC interior partition wall panel. After image processing using algorithms such as edge detection and texture analysis, a slight tear of approximately 15mm in length was identified at the tenon joint on one side of the ALC interior partition wall panel. Simultaneously, a 16-line LiDAR with a ranging accuracy of ±3cm scanned at a rate of 300,000 points per second. The generated point cloud data, after filtering and segmentation, accurately calculated the actual geometric dimensions of the ALC interior partition wall panel to be 599.2mm × 2948.7mm, and detected a very slight bend in the panel surface, with a maximum bend height of 2.1mm, which meets the national standard tolerance of ≤3mm. Through fusion processing, the six-degree-of-freedom pose of the ALC inner partition wall panel relative to the base coordinate system of the wall panel construction robot was accurately calculated. The spatial coordinates of its lower left corner point are (1250.5, -305.2, 12.8), and there is a 1.5° deviation angle between the plane of the ALC inner partition wall panel and the forward direction of the wall panel construction robot.

[0029] According to the wall panel assembly application standard, defect detection and visual positioning are performed on the current wall panel appearance image data and wall panel laser point cloud data to determine the available wall panel space location.

[0030] Furthermore, this application also includes the following steps: determining standard wall panel appearance image data according to the wall panel assembly application standard; performing defect comparison detection on the current wall panel appearance image data according to the standard wall panel appearance image data to obtain the current wall panel defect detection result; when the current wall panel defect detection result reaches a preset unqualified defect threshold, marking the current wall panel as an unqualified product and automatically skipping the current wall panel; if the current wall panel defect detection result does not reach the preset unqualified defect threshold, performing visual positioning based on the wall panel laser point cloud data to determine the available wall panel spatial location.

[0031] Furthermore, this application also includes the following steps: performing multi-dimensional feature extraction on the standard wall panel appearance image data and the current wall panel appearance image data respectively to obtain a standard wall panel appearance image feature set and a current wall panel appearance image feature set; performing loss detection comparison on the current wall panel appearance image feature set according to the standard wall panel appearance image feature set to obtain a current wall panel loss feature set; establishing a wall panel defect assessment system, and using the wall panel defect assessment system to assess the defect degree of the current wall panel loss feature set to obtain the current wall panel defect detection result.

[0032] Specifically, wall panel assembly application standards are pre-defined. Specifications regarding the quality and geometry of ALC interior partition walls are typically derived from industry standards and specific project requirements, clearly defining the acceptable limits for defects, their size, and quantity. This includes aspects such as the physical properties, appearance quality, and installation methods of the ALC interior partition walls. According to these standards, an ideal standard wall panel is defined as having no visual defects such as cracks, chipped corners, scratches, or color differences, and must meet all design and construction requirements, such as accurate dimensions, a smooth and defect-free surface, and uniform coating. The appearance image data of the standard wall panel is usually captured by a high-resolution camera or scanning equipment under defect-free conditions. The image clarity, contrast, and color information should meet requirements for subsequent processing.

[0033] Based on the wall panel assembly application standards, standard wall panel appearance image data is determined. This involves selecting a wall panel that meets the assembly standards, ensuring its appearance and dimensions conform to the standards, and that no obvious defects are present. A high-resolution camera is used to photograph the entire wall panel from a suitable distance, ensuring the image clearly captures the panel's texture, edges, and any potential minor defects. This standard wall panel appearance image data serves as a positive sample for comparison with images of the ALC interior partition wall panels collected on-site. The high-resolution camera should have sufficient lighting and a suitable angle to avoid interference from shadows and reflections. Post-processing is performed on the collected standard wall panel appearance image data to enhance image quality, such as adjusting brightness and contrast, and removing noise.

[0034] Computer vision algorithms were used to extract multidimensional features from both standard and current wall panel appearance image data, refining high-dimensional, comparable feature sets for both: the standard wall panel appearance image feature set and the current wall panel appearance image feature set. First, multidimensional features, including color, texture, shape, and edge features, were extracted from the standard wall panel appearance image data. Similarly, the same types of features, including color, texture, shape, and edge features, were extracted from the current wall panel appearance image data. After feature extraction, the standard wall panel appearance image data yielded the standard wall panel appearance image feature set. The standard wall panel appearance image data should be free of obvious defects, so all indicators in its feature set should conform to standards. The current wall panel appearance image feature set contained all actual image data. When the standard and current wall panel appearance image feature sets were compared, differences were detected and used for further defect detection and localization.

[0035] Color feature extraction typically employs color histograms, such as those in RGB or HSV color spaces, to calculate the difference in color distribution between the current wall panel and a standard wall panel. Texture feature extraction usually uses methods like the gray-level co-occurrence matrix to analyze texture directionality and roughness. Shape feature extraction typically uses shape descriptors to quantify the geometry of the wall panel, such as contour analysis. Edge detection uses edge detection methods like Canny and Sobel to identify cracks or edge variations. These feature extraction techniques are all existing and can be directly applied.

[0036] Using a similarity calculation method, the current wall panel appearance image feature set is compared and analyzed against the standard wall panel appearance image feature set to determine the difference between them. This results in the current wall panel loss feature set, which represents the defects in the current wall panel appearance image feature set relative to the standard wall panel appearance image feature set. For example, comparing the color histograms of the standard and current wall panel appearance image feature sets checks for color uniformity and color difference; comparing the gray-level co-occurrence matrices of the two sets detects texture consistency; comparing edge detection results checks for geometric consistency between the two wall panels, especially for cracks, deformation, or chipped corners; and comparing edge detection results, particularly in vulnerable areas such as tenons and joints.

[0037] By comparing the results, the degree of loss, or difference, between the two wall panels in each feature dimension is calculated. For example, measured by the cross-entropy of the color histogram, the color difference might be 5%; calculated by the similarity of texture features, i.e., the similarity value based on the gray-level co-occurrence matrix, the difference might be 0.2; if the geometry of the wall panel changes, it might introduce an error of 1mm to 2mm, or a crack, such as 15mm, might appear. The loss detection comparison results will generate a current wall panel loss feature set, representing all differences between the standard wall panel appearance image data and the current wall panel appearance image data, including color, texture, shape, edges, etc. For example, a color difference of 2%, a texture difference of 0.1%, and a shape difference of 15mm cracks mean that the current wall panel loss feature set includes color deviation, missing texture, shape deviation, and cracks.

[0038] A preset threshold for unqualified defects is used to determine the quantitative critical value of a panel as unusable. It is a decision rule based on the digitalization of wall panel assembly application standards. For example, if the crack width is ≥0.2mm and the crack length is >30mm, or the corner area is >4cm² and the depth is >5mm, or the proportion of honeycomb pitting on the panel surface is >1%, or the color difference is >5%, it is judged as unqualified.

[0039] If the current wall panel defect detection result reaches the preset non-conforming defect threshold, the current wall panel is marked as a non-conforming product and automatically skipped to ensure that non-conforming wall panels are not used in the assembly process, and the next wall panel is processed directly. For example, if the crack length of the current wall panel is 35mm and the width is 0.25mm, the defect degree of the wall panel exceeds the preset threshold, it is immediately marked as non-conforming, and the wall panel is skipped to process the next wall panel. If the current wall panel defect detection result does not reach the preset non-conforming defect threshold, the current wall panel is marked as a conforming product. For wall panels not marked as non-conforming, laser point cloud data is used to visually locate the wall panels. The laser point cloud data of the wall panels scanned by the lidar provides information such as the geometric shape, size, and position of the wall panels, from which the spatial positions of usable wall panels are determined, that is, the spatial positions of all usable wall panels are determined, including information such as the length, width, height, and angle of the wall panels, for subsequent assembly work. Through defect detection, the wall panel construction robot can automatically detect defects in the wall panels and determine whether the wall panels are qualified based on the degree of defects. This not only improves the quality control level of the wall panels, but also ensures that only qualified wall panels are used in construction, thereby improving the overall construction quality.

[0040] Furthermore, this application also includes the following steps: filtering and denoising the laser point cloud data of the wall panel and extracting key feature points to obtain a set of key feature points of the wall panel; fitting the edge contour of the set of key feature points of the wall panel according to the standard wall panel appearance image data to generate wall panel edge contour information; estimating the pose of the laser point cloud data of the wall panel based on the wall panel edge contour information to determine the wall panel placement pose information; mapping and transforming the wall panel placement pose information to the world coordinate system to determine the spatial position of the available wall panel.

[0041] Specifically, the laser point cloud data of the wall panels obtained through LiDAR scanning is filtered and denoised, and key feature points are extracted. A filtering algorithm is used to denoise the laser point cloud data, optimizing the quality of the point cloud data by removing outliers and reducing unnecessary points, ensuring that only valid spatial information is retained. Outliers, i.e., noise points, are identified and removed by analyzing the distances to the neighborhood of each point. If the distance within the neighborhood of a point is abnormally large, then that point is considered noise and removed. After filtering and denoising, irrelevant noise points in the point cloud data are removed, making the data cleaner and more accurate. For example, assuming the original point cloud data contains 1 million points, after filtering, 10% of the noise points are removed, leaving 900,000 valid points in the point cloud data.

[0042] Key feature points are extracted from the denoised point cloud data using SIFT (Scale Invariant Feature Transform). These points are typically located at the edges, corners, or other locations with significant geometric features of the wall panel, reflecting its main geometric shape and spatial distribution, such as edges, corners, or surface curvature points. The extracted key feature points are then grouped together to form a key feature point set for the wall panel, containing its main geometric features and representing its overall shape and spatial layout.

[0043] From the set of key feature points of the wall panel, feature points located at the edge of the wall panel are selected for further processing, such as the start point, end point, and turning point of the wall panel contour. An edge fitting algorithm is used to fit the selected edge feature points to obtain the edge contour of the wall panel. The optimal fitting curve is found by minimizing the distance from each point to the fitted curve. The fitting result generates the edge contour information of the wall panel, which includes not only the outer shape of the wall panel but also geometric properties such as edge smoothness, curvature, and angle. The fitting algorithm obtains the edge curve of the wall panel and calculates the curvature information of each point, thus determining whether the edge of the wall panel has obvious bending or deformation. The edge contour information of the wall panel is the geometric information of the wall panel edge obtained through the fitting algorithm, including the edge shape, length, and curvature.

[0044] Based on the edge contour information of the wall panel, a pose estimation algorithm is used to determine the position and orientation of the wall panel in three-dimensional space. Pose estimation includes the wall panel's position (X, Y, Z) and rotation angles, including roll, pitch, and yaw. For example, through edge contour fitting, the wall panel's position is obtained as (1250, 3050, 10), with rotation angles of roll = 0.5°, pitch = 0.5°, and yaw = 1.0°. Pose estimation refers to calculating the wall panel's edge contour information to estimate its position and orientation in three-dimensional space, used to determine the wall panel's precise position and orientation in space. The wall panel's placement pose information includes its position coordinates and rotation angles in three-dimensional space.

[0045] The position and orientation of the wall panel obtained through pose estimation are usually in a local coordinate system. To use this information in actual construction, it needs to be transformed from the local coordinate system to the world coordinate system using a coordinate transformation matrix. For example, assuming the wall panel position in the local coordinate system is (1250, 3050, 10), and assuming the coordinate system transformation formula between the local and world coordinate systems is: the origin of the X local coordinate system is offset by 1000mm, the origin of the Y local coordinate system is offset by -500mm, and the origin of the Z local coordinate system is offset by 5mm, after the coordinate system transformation, the position of the wall panel in the world coordinate system might be (2250, 2550, 15). After the coordinate system transformation, the spatial position of the wall panel has been mapped from the local coordinate system to the world coordinate system, determining the actual position and orientation of the wall panel on the construction site. Laser point cloud data filtering and denoising reduces noise interference, retaining 900,000 valid points, ensuring the high accuracy and reliability of the point cloud data. By estimating the pose, the position and orientation of the wall panels are accurately calculated with an error controlled within 2mm, ensuring that the wall panels can be assembled in the predetermined positions. The wall panel positions in the local coordinate system are accurately mapped to the world coordinate system, enabling the precise determination of the wall panel's position on the construction site.

[0046] Collect 3D point cloud data of the construction site, and plan the assembly path of the available wall panels based on the 3D point cloud data of the construction site to determine the assembly path of the target wall panels.

[0047] Furthermore, this application also includes the following steps: filtering and denoising the three-dimensional point cloud data of the construction site and performing spatial modeling to generate a spatial model of the wall panel construction site; marking the wall panel assembly area in the spatial model of the wall panel construction site to determine the target wall panel assembly space location; and planning the assembly path based on the available wall panel space location as the starting point and the target wall panel assembly space location as the ending point, based on the spatial model of the wall panel construction site to obtain the target wall panel assembly path.

[0048] Furthermore, this application also includes the following steps: obtaining the motion constraint parameters of the wall panel construction robot; performing obstacle identification on the wall panel construction site space model based on the motion constraint parameters to generate an available wall panel construction space model; constructing a wall panel assembly path cost function according to the internal partition wall panel assembly target; taking the available wall panel space location as the starting point and the target wall panel assembly space location as the ending point, performing path planning and solving on the wall panel assembly path cost function based on the wall panel construction site space model to obtain the target wall panel assembly path.

[0049] Specifically, LiDAR is used to collect 3D point cloud data of the construction site, which involves gathering point cloud data for the entire construction site. Point cloud data typically consists of millions of data points, each representing the 3D coordinates of a specific point on the construction site, providing detailed spatial information about structures such as wall panels, floors, walls, and ceilings. Since LiDAR scanning data is often affected by the surrounding environment, such as dust, lighting, and reflections, noise reduction processing is necessary. The purpose of noise reduction is to remove irrelevant data points caused by measurement errors or external interference, preserving valid spatial information. By dividing the point cloud data into multiple voxels and averaging the points within each voxel, redundant information is reduced.

[0050] The denoised point cloud data will be processed and converted into a 3D spatial model. This involves extracting detailed geometric information of the construction site from the point cloud data and generating a 3D model suitable for analysis, visualization, and assembly. Using surface reconstruction methods, the wall surface will be transformed into a continuous, smooth surface. Through meshing, all point cloud data will be connected to form a triangular mesh, representing the complete structure of the construction site. Using the modeled data, the generated 3D spatial model will accurately reflect the actual situation of the construction site, including the relative positions of the walls, the openings of doors and windows, and the spatial relationships with other facilities, providing all spatial constraints for wall panel assembly. In the generated wall panel construction site spatial model, the available wall panel assembly areas, i.e., the target locations for ALC interior partition wall panel installation, are marked according to the wall panel dimensions and assembly requirements.

[0051] Obtaining the motion constraint parameters of the wall panel construction robot, namely its motion capabilities and limitations, including maximum speed, turning radius, maximum load, and maximum working range, determines the robot's range of motion and operational capabilities on the construction site. Maximum speed is the robot's maximum movement speed during operation; speed limits help ensure operational stability and safety. Maximum turning radius is the minimum radius the robot can turn; typically, the robot's rotation is limited by the structure, and a too-small turning radius may lead to collisions. Maximum load is the maximum weight the robot can bear; the weight of the wall panel and the robot's load-bearing capacity directly affect its handling and assembly capabilities. Maximum working range is the robot's working area, i.e., the farthest distance the robot can reach. For example, assuming a maximum turning radius of 1.5m, a maximum load of 500kg, and a maximum speed of 0.5m / s, these parameters determine the robot's range of movement and handling capabilities on the construction site.

[0052] Obstacle identification is performed based on the spatial model of the wall panel construction site, identifying all areas and objects that may affect the robot, such as fixed structures, construction equipment, and temporary facilities. Fixed structures include columns, walls, door frames, and window frames; construction equipment includes cranes and concrete mixers; and temporary facilities include cables, wires, and pipes. Using 3D point cloud data of the construction site, all facilities can be identified. Combined with the motion constraint parameters of the wall panel construction robot, obstacles that the robot may encounter during movement, such as walls, equipment, pipes, and columns, are calculated and must be avoided to ensure the robot can successfully complete its task. For example, a 2m diameter column and a 3m high pipe are detected and marked as obstacles. By identifying and avoiding all obstacles, a series of areas suitable for wall panel assembly are defined. These areas are unaffected by obstacles and conform to the robot's motion constraints; these are the usable wall panel construction spatial models, including the free space around all obstacles and areas related to wall panel assembly. In addition to obstacle identification, the motion constraint parameters of the wall panel construction robot are also considered to ensure smooth robot movement within the generated usable space areas. For example, path planning needs to consider the robot's maximum turning radius to ensure that the robot can make appropriate turns in space without hitting obstacles. For instance, the robot needs to avoid a 2m diameter pillar and pass under a 3m high pipe while maintaining a turning radius of at least 1.5m.

[0053] Based on the installation requirements and design goals of the ALC internal partition wall panels, the parameters of the assembly target are determined, namely the internal partition wall panel assembly target, including the specific location of the wall panels, the splicing sequence, and splicing requirements. For example, the wall panels need to be installed in a left-to-right order, and the docking accuracy error between each wall panel cannot exceed 2mm. Based on the internal partition wall panel assembly target, a wall panel assembly path cost function is constructed, including distance cost, obstacle avoidance cost, robot motion constraint cost, and special requirements of the assembly target, such as the wall panel splicing sequence and accuracy requirements. Distance cost is the actual distance from the current wall panel position to the target assembly position, affecting the robot's movement distance. Due to the existence of obstacles on the construction site, bypassing obstacles requires additional movement space and time, generating additional costs, i.e., obstacle avoidance cost. Considering constraints such as the robot's turning radius and maximum load, path planning needs to avoid narrow spaces or overload operations, i.e., robot motion constraint cost. If the wall panel assembly sequence must be executed according to specific rules, or if specific docking accuracy is required, such as an error within 2mm, it also needs to be considered in path planning, increasing the corresponding cost.

[0054] In path planning, the starting point is the available wall panel space, i.e., the current location of the ALC interior partition wall panel to be installed, and the ending point is the target wall panel assembly space, i.e., the location where the ALC interior partition wall panel needs to be placed. The wall panel construction robot starts from the available wall panel space, moves the wall panel, and moves it to the target wall panel assembly space. For example, if the starting point is (10,5,1) and the ending point is (100,50,3), the actual path length is calculated as an Euclidean distance of 100.64m. If there are obstacles in the path, such as a 2m diameter pillar that needs to be bypassed, an additional 5m of distance can be added. The robot has a 1.5m turning radius, so it must avoid spaces that are not wide enough to prevent the path from becoming too narrow, which would add 5m to the path. The wall panel assembly sequence must be from left to right, and the splicing error must not exceed 2mm, which would add 2m to the path. These are the wall panel assembly path costs that need to be considered in path planning, i.e., the total path cost is 112.64m.

[0055] The algorithm uses a path planning system to select the optimal path based on the estimated cost of the current path and the distance to the target point. Initialization is performed based on the starting and ending points, obstacles, and motion constraints. A heuristic function is used to estimate the cost from the current point to the target point. In two-dimensional space, the heuristic function typically uses Euclidean distance to calculate the estimate. This estimate is combined with the cost of the current path, and the path with the minimum total cost is selected for expansion. Starting from the starting point, the algorithm expands progressively to neighboring nodes until the destination is found. During each expansion, the algorithm evaluates a node based on the following cost: f(n) = g(n) + h(n), where g(n) is the actual distance from the starting point to the current node, and h(n) is the heuristic estimate from the current node to the target node. During path search, the algorithm detects obstacles in the path and forces its way around them. For example, a 2m diameter pillar and a 3m pipe are marked as impassable areas, and the path is automatically adjusted to avoid these obstacles. After multiple expansions and evaluations, an optimal path from the starting point to the destination is generated. Assume the final path has a total length of 112.64m and has avoided all obstacles. The target wall panel assembly path is the optimal path from the available wall panel space location to the target wall panel assembly space location. It is the best movement route for the robot from the current wall panel location to the target location, taking into account all obstacles, motion constraints, and assembly requirements.

[0056] By filtering and denoising and spatial modeling, a high-precision three-dimensional construction site spatial model was generated. Taking into account the influence of obstacles, the path planning avoided all important obstacles, ensuring that the wall panel construction robot could successfully complete the wall panel assembly task. The wall panel construction robot can move smoothly in the complex construction site and complete the assembly task, ensuring the efficiency, safety and accuracy of the construction process.

[0057] The wall panel construction robot uses the force sensor to analyze the current wall panel appearance image data, determines the wall panel assembly control parameters, and controls the wall panel construction robot to perform ALC internal partition wall panel assembly based on the wall panel assembly control parameters and the target wall panel assembly path.

[0058] Furthermore, this application also includes the following steps: obtaining standard wall panel weight information from the current wall panel appearance image data; performing loss correction on the standard wall panel weight information based on the current wall panel loss feature set to obtain current wall panel weight information; determining the robot assembly contact point based on the current wall panel appearance image data; obtaining the clamping force constraint parameters of the wall panel construction robot; and performing control analysis on the robot assembly contact point based on the force sensor using the motion constraint parameters, the clamping force constraint parameters, and the current wall panel weight information to obtain wall panel assembly control parameters.

[0059] Furthermore, this application also includes the following steps: performing assembly control analysis on the robot assembly contact point based on the motion constraint parameters, the clamping force constraint parameters, and the current wall panel weight information to obtain initial assembly control parameters; the wall panel construction robot executes wall panel assembly control based on the initial assembly control parameters, while simultaneously collecting clamping force sensing data during the assembly process through the force sensor; and adjusting and optimizing the initial assembly control parameters based on the clamping force sensing data to determine the wall panel assembly control parameters.

[0060] Furthermore, this application also includes the following steps: performing steady-state over-limit detection and feedback optimization analysis on the clamping force sensing data to obtain the force feedback optimization direction; and iteratively adjusting and optimizing the initial assembly control parameters based on the force feedback optimization direction to determine the wall panel assembly control parameters.

[0061] Specifically, based on production specifications, the standard wall panel weight information obtained from the current wall panel appearance image data is typically a preset value, taking into account factors such as the wall panel's size and material density. For example, if the standard wall panel's dimensions are 600mm × 2950mm × 100mm and its material density is 1.5g / cm³, then the standard wall panel's weight is usually 267kg. Based on the current wall panel's loss feature set, it is determined that the current wall panel has a 10mm crack and a 1cm² missing corner. These defects will cause the wall panel's actual weight to be reduced. The wall panel's weight is then corrected according to the loss feature set, i.e., the weight loss caused by the current wall panel's loss feature set is subtracted, to obtain the current wall panel's weight information, i.e., the actual weight of the current wall panel.

[0062] Based on the current wall panel appearance image data, the robot assembly contact points are determined, which are the areas where the wall panel construction robot contacts the ALC internal partition wall panel. These are typically located at the edge of the wall panel for clamping and positioning. The determination of the contact points is based on the geometric features of the ALC internal partition wall panel (such as size and shape) and the operating range of the wall panel construction robot arm. For example, the four corners of the wall panel are often used as contact points to ensure the stability of clamping and handling. The clamping force constraint parameter of the wall panel construction robot is the maximum clamping force that the wall panel construction robot can apply during operation. This is used to ensure that the robot does not damage the wall panel when clamping it, while also ensuring the stability of the wall panel. For example, assuming that the maximum clamping force that the wall panel construction robot can apply is 500N, it means that no force exceeding 500N will be applied during the assembly task to prevent damage to the wall panel.

[0063] Based on the weight and structural characteristics of the wall panel, a suitable initial clamping force is calculated. Based on the wall panel's weight and motion constraints, the appropriate assembly position and adjustment angle are calculated to ensure precise placement of the wall panel at the target location, and suitable assembly control parameters are calculated. By analyzing the position of the assembly contact points, the clamping force requirements for each contact point are determined to ensure that the wall panel does not slip or cause errors in assembly accuracy. Assembly control analysis refers to calculating appropriate assembly control parameters, including clamping force, position, and posture adjustment, based on motion constraint parameters, clamping force constraint parameters, and current wall panel weight information, to ensure accuracy and safety during the assembly process. The initial assembly control parameters, obtained after assembly control analysis, serve as the basis for the wall panel construction robot to perform assembly tasks, including control parameters such as clamping force, assembly speed, and posture adjustment.

[0064] The wall panel construction robot begins assembling the wall panels according to the initial assembly control parameters. For example, the robot's clamping force is set to 250N, and it moves to the target position to ensure the wall panel is securely fixed while avoiding over-clamping that could damage it. During assembly, force sensors continuously monitor changes in the clamping force, including the current clamping force value and the rate of force change. For instance, if the force sensor detects a slight adjustment in the clamping force from 250N to 245N, it indicates that the wall panel is stable and the clamping force has not exceeded the safety limit. The real-time clamping force data collected by the force sensor helps detect whether the preset safety range has been exceeded and adjusts the assembly strategy as needed.

[0065] Steady-state over-limit detection is performed on the real-time collected clamping force sensing data. If the clamping force exceeds the robot's maximum clamping force, an alarm is immediately issued, and the assembly control strategy is adjusted. Assuming the clamping force sensing data exhibits some fluctuation, these fluctuations are analyzed to optimize clamping force control. For example, if the clamping force is too low, it may lead to an unstable wall panel, so the clamping force is increased; if the clamping force is too high, it is reduced to prevent damage to the wall panel. Furthermore, if large fluctuations in clamping force are detected, feedback analysis indicates that uneven pressure is caused by uneven surface of the wall panel; adjusting the clamping force or changing the clamping point ensures stable installation.

[0066] In other words, steady-state over-limit detection involves steady-state monitoring of real-time clamping force sensing data to detect whether the clamping force exceeds a preset threshold. If the clamping force exceeds the robot's maximum tolerance or the wall panel's tolerance limit, appropriate measures are taken. This involves not only detecting the instantaneous clamping force but also its stability. For example, if the clamping force fluctuates continuously outside a certain range, or if the clamping force suddenly increases or decreases within a short period, it will also be identified as abnormal.

[0067] Based on the steady-state detection results, feedback optimization analysis is performed to determine whether the current clamping force control is appropriate. Based on real-time clamping force sensing data, it is determined whether the clamping force needs adjustment or the assembly method needs to be changed. For example, too low a clamping force may lead to wall panel instability, while too high a clamping force may lead to wall panel deformation or damage. If the clamping force is maintained at 240N and the wall panel does not show instability or deformation, the clamping force is appropriate. If the clamping force remains at 220N or below, the wall panel may not be sufficiently clamped, requiring an increase in clamping force. Based on the feedback analysis results, the direction of adjustment is determined. If the clamping force is detected to be too low or uneven, the clamping force needs to be increased; if the clamping force is too high and causes wall panel deformation or damage, a decision will be made to reduce the clamping force. This ensures that the clamping force is maintained within a reasonable range, avoiding instability or damage caused by excessively high or low clamping forces. The force feedback optimization direction, based on the feedback optimization analysis, determines the direction for adjusting the clamping force or control strategy, meaning increasing, decreasing, or redistributing the clamping force to achieve the best assembly effect.

[0068] The initial assembly control parameters are gradually adjusted based on force feedback optimization. This means that the control strategy is progressively optimized based on the results of each iteration to ensure the safe and stable assembly of the wall panels. Specifically, based on feedback analysis, the initial assembly control parameters are adjusted, clamping force data is re-collected, and the expected assembly effect is checked. Based on the latest feedback results, the parameters are analyzed and optimized again. If the clamping force still does not meet the requirements, adjustments continue until the optimal state is reached. After multiple iterations and adjustments, the final control parameters suitable for the wall panel assembly process are determined. The wall panel assembly control parameters are used to control various parameters during the wall panel assembly process, including clamping force, assembly speed, and posture adjustment, which determine the stability and accuracy of the assembly process.

[0069] The wall panel assembly control parameters and the target wall panel assembly path are sent to the control system of the wall panel construction robot. Based on these parameters and the target path, the robot begins the assembly task. The robot travels along the planned path, performing multiple operations including clamping, positioning, and installation. During assembly, the robot uses clamping force constraints to ensure the force remains within a set range, based on the target path and control parameters. By adjusting the clamping force in real time, the robot ensures the wall panels are not damaged by excessive force or become unstable due to insufficient force during transport. During movement, the robot adjusts its posture according to the actual position and target assembly location of the wall panels to ensure precise placement. Sensors monitor the accuracy of the wall panels in real time to ensure error-free joints and compliance with assembly requirements. During the assembly process, the wall panel construction robot monitors the clamping force, position, and path accuracy through built-in sensors, and adjusts the control parameters of the wall panel construction robot in real time to ensure the smooth progress of the assembly task.

[0070] Once the wall panel construction robot has moved from the starting point to the target assembly position and successfully placed the wall panel in the designated location, the assembly process is confirmed to be complete. Sensors verify the assembly accuracy, ensuring there are no obvious gaps between the wall panels and that the panel alignment meets design requirements. For example, the wall panel is successfully moved from the starting point to the target assembly position, and the assembly error is confirmed to be within 2mm. By precisely controlling the wall panel assembly path and control parameters, the wall panel construction robot can efficiently complete the handling and assembly tasks of the wall panels, reducing manual intervention, increasing assembly efficiency by 30%, avoiding safety accidents caused by excessive clamping force or path errors, and ensuring a safe distance between the wall panel construction robot, the wall panels, and the environment during construction.

[0071] In summary, the ALC internal partition wall panel assembly control method combined with machine vision provided in this application has the following beneficial effects: A wall panel construction robot, integrating a binocular camera, LiDAR, and force sensor, is acquired. The binocular camera and LiDAR collect current wall panel appearance image data and wall panel laser point cloud data. Defect detection and visual positioning are performed on the current wall panel appearance image data and wall panel laser point cloud data according to wall panel assembly application standards to determine the available wall panel spatial location. Three-dimensional point cloud data of the construction site is collected, and assembly path planning is performed on the available wall panel spatial location based on the three-dimensional point cloud data of the construction site to determine the target wall panel assembly path. The wall panel construction robot performs control analysis on the current wall panel appearance image data based on the force sensor to determine wall panel assembly control parameters. Based on the wall panel assembly control parameters and the target wall panel assembly path, the wall panel construction robot is controlled to perform ALC internal partition wall panel assembly control. In other words, data acquired using binocular cameras and lidar is used to determine whether there are quality problems with the wall panels through defect detection and visual positioning, and to determine whether the wall panels can be used for assembly. The three-dimensional point cloud data of the construction site is used to plan the assembly path of the wall panels, ensuring that the spatial layout during the assembly process meets the design requirements. Based on the feedback from force sensors and visual data, assembly control parameters are formulated to control the wall panel construction robot to accurately assemble the wall panels, avoid deviations or damage, reduce assembly errors, and improve the assembly quality of the ALC interior partition wall panels.

[0072] Example 2: Based on the same inventive concept as the ALC internal partition panel assembly control method combined with machine vision in Example 1, this application also provides an ALC internal partition panel assembly control system combined with machine vision. Please refer to the appendix. Figure 2 The ALC interior partition panel assembly control system incorporating machine vision includes: The robot identification module 11 is used to acquire the wall panel construction robot, which integrates a binocular camera, a lidar, and a force sensor. The binocular camera and lidar acquire current wall panel appearance image data and wall panel laser point cloud data. The vision inspection module 12 is used to perform defect detection and visual positioning on the current wall panel appearance image data and wall panel laser point cloud data according to wall panel assembly application standards, determining the available wall panel spatial location. The path planning module 13 is used to acquire three-dimensional point cloud data of the construction site, and based on the three-dimensional point cloud data of the construction site, to plan the assembly path for the available wall panel spatial location, determining the target wall panel assembly path. The assembly control module 14 is used to control and analyze the current wall panel appearance image data based on the force sensor, determine wall panel assembly control parameters, and control the wall panel construction robot to perform ALC internal partition wall panel assembly control based on the wall panel assembly control parameters and the target wall panel assembly path.

[0073] Furthermore, the vision inspection module 12 in the ALC interior partition wall panel assembly control system combined with machine vision is also used for: determining standard wall panel appearance image data according to the wall panel assembly application standard; performing defect comparison and detection on the current wall panel appearance image data according to the standard wall panel appearance image data to obtain the current wall panel defect detection result; when the current wall panel defect detection result reaches the preset unqualified defect threshold, marking the current wall panel as unqualified and automatically skipping the current wall panel; if the current wall panel defect detection result does not reach the preset unqualified defect threshold, performing visual positioning based on the wall panel laser point cloud data to determine the available wall panel spatial position.

[0074] Furthermore, the vision detection module 12 in the ALC interior partition wall panel assembly control system combined with machine vision is also used for: performing multi-dimensional feature extraction on the standard wall panel appearance image data and the current wall panel appearance image data respectively to obtain the standard wall panel appearance image feature set and the current wall panel appearance image feature set; performing loss detection comparison on the current wall panel appearance image feature set according to the standard wall panel appearance image feature set to obtain the current wall panel loss feature set; building a wall panel defect assessment system, and using the wall panel defect assessment system to assess the defect degree of the current wall panel loss feature set to obtain the current wall panel defect detection result.

[0075] Furthermore, the vision detection module 12 in the ALC internal partition wall panel assembly control system combined with machine vision is also used for: filtering and denoising the laser point cloud data of the wall panel and extracting key feature points to obtain a set of key feature points of the wall panel; fitting the edge contour of the set of key feature points of the wall panel according to the standard wall panel appearance image data to generate wall panel edge contour information; estimating the pose of the laser point cloud data of the wall panel based on the wall panel edge contour information to determine the wall panel placement pose information; mapping and transforming the wall panel placement pose information to the world coordinate system to determine the spatial position of the available wall panel.

[0076] Furthermore, the path planning module 13 in the ALC interior partition wall panel assembly control system combined with machine vision is also used for: filtering and denoising the 3D point cloud data of the construction site and spatial modeling to generate a spatial model of the wall panel construction site; marking the wall panel assembly area in the spatial model of the wall panel construction site to determine the target wall panel assembly space location; and planning the assembly path based on the available wall panel space location as the starting point and the target wall panel assembly space location as the ending point, based on the spatial model of the wall panel construction site to obtain the target wall panel assembly path.

[0077] Furthermore, the path planning module 13 in the ALC internal partition wall panel assembly control system combined with machine vision is also used to: acquire the motion constraint parameters of the wall panel construction robot; perform obstacle recognition on the wall panel construction site space model based on the motion constraint parameters to generate an available wall panel construction space model; construct a wall panel assembly path cost function according to the internal partition wall panel assembly target; take the available wall panel space location as the starting point and the target wall panel assembly space location as the ending point, and perform path planning and solution on the wall panel assembly path cost function based on the wall panel construction site space model to obtain the target wall panel assembly path.

[0078] Furthermore, the assembly control module 14 in the ALC interior partition wall panel assembly control system combined with machine vision is also used for: acquiring standard wall panel weight information of the current wall panel appearance image data; performing loss correction on the standard wall panel weight information based on the current wall panel loss feature set to obtain the current wall panel weight information; determining the robot assembly contact point according to the current wall panel appearance image data; acquiring the clamping force constraint parameters of the wall panel construction robot; and performing control analysis on the robot assembly contact point based on the force sensor using the motion constraint parameters, the clamping force constraint parameters, and the current wall panel weight information to obtain wall panel assembly control parameters.

[0079] Furthermore, the assembly control module 14 in the ALC internal partition wall panel assembly control system combined with machine vision is also used to: analyze the assembly control of the robot assembly contact point based on the motion constraint parameters, the clamping force constraint parameters, and the current wall panel weight information to obtain initial assembly control parameters; the wall panel construction robot executes wall panel assembly control based on the initial assembly control parameters, and simultaneously collects clamping force sensing data during the assembly process through the force sensor; and adjusts and optimizes the initial assembly control parameters based on the clamping force sensing data to determine the wall panel assembly control parameters.

[0080] Furthermore, the assembly control module 14 in the ALC internal partition wall panel assembly control system combined with machine vision is also used for: performing steady-state over-limit detection and feedback optimization analysis on the clamping force sensing data to obtain the force feedback optimization direction; and iteratively adjusting and optimizing the initial assembly control parameters based on the force feedback optimization direction to determine the wall panel assembly control parameters.

[0081] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Figure 1The ALC internal partition panel assembly control method and specific examples combining machine vision in Embodiment 1 are also applicable to the ALC internal partition panel assembly control system combining machine vision in this embodiment. Through the foregoing detailed description of the ALC internal partition panel assembly control method combining machine vision, those skilled in the art can clearly understand the ALC internal partition panel assembly control system combining machine vision in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0082] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0083] Obviously, those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. An ALC internal partition wall panel assembly control method combining machine vision, characterized in that, include: A wall panel construction robot is acquired. The wall panel construction robot integrates a binocular camera, a lidar, and a force sensor. The binocular camera and lidar are used to collect current wall panel appearance image data and wall panel laser point cloud data. According to the wall panel assembly application standard, defect detection and visual positioning are performed on the current wall panel appearance image data and wall panel laser point cloud data to determine the available wall panel space location. Collect 3D point cloud data of the construction site, and plan the assembly path of the available wall panels based on the 3D point cloud data of the construction site to determine the target wall panel assembly path. The wall panel construction robot uses the force sensor to analyze the current wall panel appearance image data, determines the wall panel assembly control parameters, and controls the wall panel construction robot to perform ALC internal partition wall panel assembly based on the wall panel assembly control parameters and the target wall panel assembly path.

2. The ALC interior partition panel assembly control method combined with machine vision as described in claim 1, characterized in that, Determining the available wall panel space includes: Based on the wall panel assembly application standards, determine the standard wall panel appearance image data; Defects are compared and detected in the current wall panel appearance image data according to the standard wall panel appearance image data to obtain the current wall panel defect detection result. When the current wall panel defect detection result reaches the preset unqualified defect threshold, the current wall panel is marked as unqualified and the current wall panel is automatically skipped. If the current wall panel defect detection result does not reach the preset unqualified defect threshold, visual positioning is performed based on the wall panel laser point cloud data to determine the available wall panel space location.

3. The ALC interior partition panel assembly control method combined with machine vision as described in claim 2, characterized in that, The process of obtaining the current wall panel defect detection results includes: Multidimensional feature extraction is performed on the standard wall panel appearance image data and the current wall panel appearance image data respectively to obtain the standard wall panel appearance image feature set and the current wall panel appearance image feature set; The current wall panel appearance image feature set is compared with the standard wall panel appearance image feature set to obtain the current wall panel loss feature set. A wall panel defect assessment system is established, and the defect severity of the current wall panel loss feature set is assessed using the wall panel defect assessment system to obtain the current wall panel defect detection result.

4. The ALC interior partition panel assembly control method combined with machine vision as described in claim 2, characterized in that, The step of visually locating the available wall panel space based on the laser point cloud data of the wall panel includes: The laser point cloud data of the wall panel is filtered and denoised, and key feature points are extracted to obtain a set of key feature points of the wall panel. The wall panel edge contour information is generated by fitting the key feature point set of the wall panel with the standard wall panel appearance image data. Based on the edge contour information of the wall panel, the pose of the laser point cloud data of the wall panel is estimated to determine the placement pose information of the wall panel. The placement pose information of the wall panel is then mapped and transformed to the world coordinate system to determine the spatial position of the available wall panel.

5. The ALC interior partition panel assembly control method combined with machine vision as described in claim 3, characterized in that, Determining the target wall panel assembly path includes: Based on the three-dimensional point cloud data of the construction site, filtering and noise reduction and spatial modeling are performed to generate a spatial model of the wall panel construction site. Mark the wall panel assembly area in the space model of the wall panel construction site to determine the target wall panel assembly space location; Based on the spatial model of the wall panel construction site, the available wall panel space location is used as the starting point and the target wall panel assembly space location is used as the ending point to plan the assembly path and obtain the target wall panel assembly path.

6. The ALC interior partition panel assembly control method combined with machine vision as described in claim 5, characterized in that, The process of obtaining the target wall panel assembly path includes: Obtain the motion constraint parameters of the wall panel construction robot, and perform obstacle recognition on the wall panel construction site space model based on the motion constraint parameters to generate a usable wall panel construction space model. Based on the internal partition wall panel assembly target, construct the wall panel assembly path cost function; Using the available wall panel space as the starting point and the target wall panel assembly space as the ending point, the path planning solution is performed on the wall panel assembly path cost function based on the wall panel construction site space model to obtain the target wall panel assembly path.

7. The ALC interior partition panel assembly control method combined with machine vision as described in claim 6, characterized in that, The determination of wall panel assembly control parameters includes: Obtain the standard wall panel weight information of the current wall panel appearance image data, and perform loss correction on the standard wall panel weight information based on the current wall panel loss feature set to obtain the current wall panel weight information; Based on the current wall panel appearance image data, determine the robot assembly contact point; The clamping force constraint parameters of the wall panel construction robot are obtained. Based on the force sensor, the robot assembly contact point is controlled and analyzed using the motion constraint parameters, the clamping force constraint parameters, and the current wall panel weight information to obtain the wall panel assembly control parameters.

8. The ALC interior partition panel assembly control method combined with machine vision as described in claim 7, characterized in that, The obtained wall panel assembly control parameters include: Based on the motion constraint parameters, the clamping force constraint parameters, and the current wall panel weight information, the robot assembly contact point is analyzed for assembly control to obtain initial assembly control parameters. The wall panel construction robot performs wall panel assembly control based on the initial assembly control parameters, and at the same time collects clamping force sensing data during the assembly process through the force sensor; Based on the clamping force sensing data, the initial assembly control parameters are adjusted and optimized to determine the wall panel assembly control parameters.

9. The ALC interior partition panel assembly control method combined with machine vision as described in claim 8, characterized in that, The determination of wall panel assembly control parameters includes: The clamping force sensing data is subjected to steady-state over-limit detection and feedback optimization analysis to obtain the force feedback optimization direction; Based on the force feedback optimization direction, the initial assembly control parameters are iteratively adjusted and optimized to determine the wall panel assembly control parameters.

10. An ALC internal partition wall panel assembly control system incorporating machine vision, characterized in that: The steps for implementing the ALC internal partition panel assembly control method incorporating machine vision as described in any one of claims 1 to 9, wherein the ALC internal partition panel assembly control system incorporating machine vision includes: The robot identification module is used to acquire the wall panel construction robot, which integrates a binocular camera, a lidar and a force sensor. The binocular camera and lidar are used to collect current wall panel appearance image data and wall panel laser point cloud data. The visual inspection module is used to perform defect detection and visual positioning on the current wall panel appearance image data and wall panel laser point cloud data according to the wall panel assembly application standard, and to determine the available wall panel space location. The path planning module is used to collect three-dimensional point cloud data of the construction site, and to plan the assembly path of the available wall panels based on the three-dimensional point cloud data of the construction site to determine the target wall panel assembly path. The assembly control module is used to control and analyze the current wall panel appearance image data based on the force sensor through the wall panel construction robot, determine the wall panel assembly control parameters, and control the wall panel construction robot to perform ALC internal partition wall panel assembly control based on the wall panel assembly control parameters and the target wall panel assembly path.

Images