Deep learning-based point cloud prefabricated wall size detection method

Abstract

The invention discloses a point cloud prefabricated wall size detection method based on deep learning, and the method comprises the steps: firstly carrying out the point cloud data collection of a structural surface of a to-be-detected wall prefabricated part, and constructing a point cloud image data set; then preprocessing the point cloud data to obtain preprocessed point cloud data; separating the point cloud data of the wall prefabricated part from the point cloud image by using a grid-based point cloud segmentation algorithm; secondly, constructing a PointNet model, extracting feature points of the boundary of the wall prefabricated part, obtaining a clustering cluster of the boundary part through a DBSCAN density clustering algorithm, and calculating the length and width of the wall prefabricated part according to clustering cluster information; and finally, in the point cloud image, visualizing the point cloud by using a visualization tool, and displaying the calculated length and width data of the wall prefabricated part on the point cloud image. According to the invention, through fusion of the deep learning algorithm and the point cloud data processing technology, automatic and high-precision detection of the wall size is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of construction engineering quality inspection, and specifically to a point cloud prefabricated wall size inspection method based on deep learning. Background Art

[0002] With the rapid development of prefabricated building technology, prefabricated walls, as a crucial component of building structures, face a significant impact on overall safety and construction efficiency, particularly regarding their dimensional accuracy and quality control. Traditional wall dimension detection methods rely primarily on manual measurement or simple measuring tools, which present significant challenges. First, manual measurement is inefficient and difficult to meet the demands of rapid construction in large-scale prefabricated buildings. Second, measurement accuracy is limited and susceptible to human factors and environmental conditions, leading to inaccurate data. Third, traditional methods struggle to adapt to the dimensional detection of complex wall structures, particularly when walls contain special features such as holes and bumps. Measurement becomes significantly more difficult, making the reliability of the results difficult to guarantee.

[0003] In modern construction projects, accurate wall dimensional measurement is crucial for ensuring construction quality, optimizing material usage, and improving efficiency. For example, in prefabricated wall production, dimensional deviations can prevent accurate component installation, increasing construction costs and time. On the construction site, rapid wall dimensional measurement can promptly identify damage during transportation or installation, preventing subsequent construction issues. Therefore, developing an efficient, accurate, and adaptable wall dimensional measurement method has become a pressing technical challenge in the field of prefabricated construction. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a point cloud prefabricated wall size detection method based on deep learning, which realizes automated and high-precision detection of wall size by integrating deep learning algorithm and point cloud data processing technology.

[0005] The technical solution of the present invention is: A method for detecting the size of prefabricated walls using point clouds based on deep learning, comprising the following steps: (1) Collect point cloud data of the structural surface of the prefabricated wall components to be tested and construct a point cloud image dataset; (2) Preprocess the point cloud data in the point cloud image to remove noise points and redundant data points to obtain preprocessed point cloud data; (3) Use the grid-based point cloud segmentation algorithm to separate the point cloud data of the wall prefabricated components from the point cloud image; (4) Construct a PointNet model to extract the feature points of the boundary of the prefabricated wall components, and obtain the clusters of the boundary of the prefabricated wall components through the DBSCAN density clustering algorithm. Calculate the length and width of the prefabricated wall components based on the cluster information; (5) In the point cloud image of the prefabricated wall component, a visualization tool is used to visualize the point cloud, and the calculated length and width data of the prefabricated wall component are displayed on the point cloud image.

[0006] The point cloud data of the structural surface of the prefabricated wall component to be measured is collected by using a 3D laser camera.

[0007] The specific steps of preprocessing the point cloud data are as follows: S21, downsampling processing: select the set radius The point cloud space is divided into two parts. The point closest to the center of each sphere is selected as the sampling point to replace all the points in the sphere. Only the sampling point of each sphere is retained in the point cloud, and the position of the sampling point does not move. S22, denoising: for each of the remaining sampling points, search its neighbor point set within a given radius. When the number of neighbor points in the neighbor point set is less than the set minimum neighbor number threshold, mark the point as an outlier and remove it from the sampling points. S23, normalization processing: normalize the remaining sampling points after denoising to eliminate the scale difference of the point cloud data so that point cloud data of different sizes have the same scale.

[0008] The specific steps of the downsampling process are: S211, select the center point of the sphere, randomly select an initial point from the point cloud as the center point of the sphere , and then search and confirm the points in the sphere, that is, calculate all the points in the point cloud and the center point The distance is not greater than the set radius The points are included in the current sphere, as shown in the following formula (1): (1); In formula (1), Represents the set of points within the current sphere, Represents a point within the current sphere, Represents a point in the point cloud, represents the center point of the sphere, represents the Euclidean distance, Represents the set radius; S212. The set of points inside the current sphere , find the distance from the center point The nearest point , as the sampling point of the current sphere, use the sampling point to replace other points in the current sphere, the position of the sampling point does not move, the sampling point The selection process is shown in the following formula (2): (2); In formula (2), Represents any point in the current sphere except the center point. represents the center point of the sphere, represents the Euclidean distance; S213. Update the point cloud and sphere center: remove all points in the searched sphere, randomly select a new sphere center point from the remaining points in the point cloud, repeat steps S21 and S22 to search for points in the sphere, and update the sampling points of the sphere until all points in the point cloud data are searched and processed or the required number of sampling points is reached.

[0009] The specific steps of the denoising process are: S221, for each sampling point , calculate its The set of neighbor points in , see the following formula (3) for details: (3); In formula (3), Representative sampling points and sampling points The Euclidean distance between S222, when the sampling point The set of neighbor points The number of neighbor points in the network is less than the set minimum number of neighbors. When , the sampling point is removed from the sampling point set .

[0010] The specific steps of the normalization process are: S231. Calculate the geometric center of the point cloud, as shown in the following formula (4): (4); In formula (4), Represents the geometric center of the point cloud composed of the remaining sampling points after denoising, Represents the total number of remaining sampling points after denoising, is the coordinate of the i-th sampling point in the point cloud composed of the remaining sampling points after denoising; S232, translate the point cloud to a position with the geometric center as the origin, and update the coordinate data of the point cloud, as shown in the following formula (5): (5); In formula (5), Represents the coordinates of the sampling point after translation, Represents the sampling coordinates before translation ; S233, calculate the maximum distance from all denoised sampling points to the origin after translation , see the following formula (6) for details: (6); S234. Scale the point cloud to the unit sphere so that the maximum distance between all points in the point cloud is 1. See the following formula (7): (7); In formula (7), Represents the normalized point coordinates.

[0011] The specific steps of using the grid-based point cloud segmentation algorithm to separate the point cloud data of the prefabricated wall components from the point cloud image are as follows: S31. Define the grid size and then perform grid mapping, as shown in the following formula (8): (8); In formula (8), is the coordinate of any point in the preprocessed point cloud data after grid mapping, is the coordinate of any point in the preprocessed point cloud data, 、 、 is the three-dimensional size of the grid; S32. Calculate the geometric features of the point cloud within each grid cell, including the normal vector and curvature, to provide the necessary geometric information for subsequent segmentation. The directional consistency of the normal vector is the key basis for determining whether the point clouds belong to the same area. The calculation of the normal vector is shown in the following formula (9): (9); In formula (9), represents the normal vector of the i-th grid cell, represents the coordinates of the points within the grid cell, represents the coordinates of the center point of the grid cell, Represents the number of points within a grid cell; The curvature calculation first calculates the local covariance matrix of the i-th grid cell , see the following formula (10) for details: (10); In formula (10), Represents the transpose operation; Then calculate the covariance matrix The eigenvalue of 、 and ,and , and then calculate the curvature according to formula (11) ; (11); S33. Calculate the data density of the point cloud in each grid cell, as shown in the following formula (12): (12); In formula (12), represents the point cloud density of the i-th grid cell, represents the volume of the i-th grid cell; S34. Setting threshold and , the data density The grid cells within the threshold range are divided into wall point clouds, as shown in the following formula (13): (13); In formula (13), Represents the wall point cloud; S35. Segmentation based on geometric features: further refine the segmentation results based on the normal vector and curvature geometric features to remove the mis-segmented point clouds in the background. The specific steps are as follows: Based on formula (9), the normal vectors of all grid cells are calculated and screened according to the direction consistency of the normal vectors. The grid cells with inconsistent normal vector directions are screened and removed. Calculate the unit normal vector of any two points within a single grid cell in the remaining grid cells and The angle between , see the following formula (14) for details: (14); In formula (14), and is the unit normal vector between any two points within a single grid cell, Represents the modulus of the vector; Set the maximum normal vector angle of the precast wall component structure surface , when the angle between all normal vectors in the grid cell is No more than , it is preliminarily determined that the grid unit belongs to the structural surface of the prefabricated wall component, otherwise the point cloud data in the grid unit is filtered and removed; Then perform curvature judgment and set the curvature threshold , when the ith grid cell calculated by equation (11) No greater than , it is finally determined that the grid unit belongs to the structural surface of the prefabricated wall component, that is, the point cloud data of the prefabricated wall component is separated from the point cloud image; otherwise, the point cloud data in the grid unit is filtered and removed, and the point cloud data in the remaining grid units are the point cloud data of the prefabricated wall component.

[0012] The steps for calculating the unit normal vector of a point within a single grid cell are: First, select any point within a single grid cell , construct its neighborhood point set : ; Calculate the neighborhood point set Center point , see the following formula (15) for details: (15); In formula (15), Representative Points The neighborhood points are the distance points in the point cloud data nearest point; Represents a set of neighborhood points The total number of neighborhood points in the Then the center point , neighborhood point set Substitute all the neighboring points in into formula (10) to calculate the neighboring point set The local covariance matrix of Replace the formula (10) , Replace the formula (10) , Replace the formula (10) ; Then calculate the neighborhood point set The eigenvalues ​​of the local covariance matrix 、 and ,and , 、 and Corresponding to three eigenvectors 、 and , the smallest eigenvalue The corresponding eigenvector That is the point The normal vector of point , is normalized according to the following formula (16) to obtain The unit normal vector ; (16); In formula (16), Yes The corresponding unit normal vector, Represents the vector modulus.

[0013] The specific steps of constructing the PointNet model, extracting the feature points of the boundary of the prefabricated wall components, and obtaining the clusters of the boundary of the prefabricated wall components through the DBSCAN density clustering algorithm, and calculating the length and width of the prefabricated wall components based on the cluster information are as follows: S41. Build a PointNet model. The PointNet model includes an input layer, a feature extraction layer, a global feature layer, and an output layer. The input layer is used to input the point cloud data of prefabricated wall components. The three-dimensional coordinate information of each point is ; The feature extraction layer extracts features from each point through multiple fully connected layers and ReLU activation functions. The initial feature of the mth point in the wall prefabricated component point cloud data is , after being processed by the fully connected layer and ReLU activation function of the jth layer, the new feature is obtained ,in and are the weights and biases of the jth layer in the feature extraction layer; The global feature layer aggregates the features of all points through the maximum pooling operation to obtain the global feature , Represents the local features of each point, is the index of the last layer of feature extraction layer; The output layer is used to convert the global features After passing through the fully connected layer, the final result is output, that is, the global features are first Broadcast to each point, and then combine with the local features of each point Splicing to form the comprehensive features of each point , Represents splicing, and then the probability of each point belonging to a boundary point is calculated through the fully connected layer and sigmoid activation function , see the following formula (17) for details; (17); In formula (17), and are the weights and biases of the fully connected layers in the output layer, Represents the sigmoid activation function; S42. Set a probability threshold ,Will Greater than The points are identified as feature points of the boundary of prefabricated wall components; S43, using the DBSCAN density clustering algorithm to obtain the clustering clusters of the boundary part of the wall prefabricated components, first set the neighborhood radius And the minimum number of points minPts, then the feature points of the identified wall prefabricated component boundary are input into the DBSCAN density clustering algorithm, and the feature point set of the wall prefabricated component boundary is , feature points The coordinates are ,The DDBSCAN density clustering algorithm is used to divide the feature points of the boundary of the wall prefabricated components into multiple clusters, each cluster represents a boundary part of the wall; S44. The feature point set of a boundary part of the wall is , feature points The coordinates are , calculate the centroid coordinates of each boundary segment , see the following formula (18) for details: (18); In formula (18), Representative point set Total number of midpoints; S45. Calculate the length and width of the wall: The centroids of the left and right boundary parts of the wall are and ,but and The coordinates of and The centroids of the upper and lower boundary parts of the wall are and ,but and The coordinates of and ; then the length of the wall and width Calculated by the following formula (19) and formula (20) respectively; (19); (20).

[0014] Advantages of the present invention: (1) The present invention preprocesses the collected point cloud data, specifically including downsampling, denoising and normalization processing. Through the orderly arrangement of the above preprocessing processes, the point cloud image data set is cleaned, noise points and redundant data are reduced, and the efficiency of subsequent point cloud data extraction and clustering is improved.

[0015] (2) The present invention uses a grid-based point cloud segmentation algorithm to separate the point cloud data of the prefabricated wall components from the point cloud image, which can effectively separate the point cloud data of the prefabricated wall components from the background, providing a basis for subsequent geometric feature extraction and size measurement.

[0016] (3) The present invention constructs a PointNet model to extract feature points of the boundary of prefabricated wall components, and obtains the boundary points of prefabricated wall components through the DBSCAN density clustering algorithm. The PointNet model can directly learn global features from point cloud data. Specifically, it extracts features from each point independently and then obtains global features through the maximum pooling operation. The DBSCAN density clustering algorithm can automatically determine the number of clusters and has good robustness to noise points.

[0017] (4) The present invention can calculate the length and width of the prefabricated wall components based on the coordinates of the boundary points of the prefabricated wall components, and visualize them on the point cloud image, intuitively presenting the dimensional data of the wall, making it easier for construction workers to complete the quality control of the wall based on the dimensional data of the wall. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0020] See Figure 1 A method for detecting the size of prefabricated walls using point clouds based on deep learning is proposed, which specifically includes the following steps: (1) Use a 3D laser camera to collect point cloud data on the structural surface of the prefabricated wall component to be measured, and construct a point cloud image dataset; (2) Preprocess the point cloud data in the point cloud image to remove noise points and redundant data points to obtain preprocessed point cloud data; The specific steps for preprocessing point cloud data are as follows: S21, downsampling processing, specifically: S211, select the center point of the sphere, randomly select an initial point from the point cloud as the center point of the sphere , and then search and confirm the points in the sphere, that is, calculate all the points in the point cloud and the center point The distance is not greater than the set radius The points are included in the current sphere, as shown in the following formula (1): (1); In formula (1), Represents the set of points within the current sphere, Represents a point within the current sphere, Represents a point in the point cloud, represents the center point of the sphere, represents the Euclidean distance, Represents the set radius; S212. The set of points inside the current sphere , find the distance from the center point The nearest point , as the sampling point of the current sphere, use the sampling point to replace other points in the current sphere, the position of the sampling point does not move, the sampling point The selection process is shown in the following formula (2): (2); In formula (2), Represents any point in the current sphere except the center point. represents the center point of the sphere, represents the Euclidean distance; S213, update point cloud and sphere center: remove all points in the searched sphere, randomly select a new sphere center point from the remaining points in the point cloud, repeat steps S21 and S22 to search for points in the sphere, and update the sampling points of the sphere until all points in the point cloud data are searched and processed or the required number of sampling points is reached S22, denoising processing, specifically: S221, for each sampling point , calculate its The set of neighbor points in , see the following formula (3) for details: (3); In formula (3), Representative sampling points and sampling points The Euclidean distance between S222, when the sampling point The set of neighbor points The number of neighbor points in the network is less than the set minimum number of neighbors. When , the sampling point is removed from the sampling point set ; S23, normalization processing: normalize the remaining sampling points after denoising to eliminate the scale difference of the point cloud data so that point cloud data of different sizes have the same scale. Specifically: S231. Calculate the geometric center of the point cloud, as shown in the following formula (4): (4); In formula (4), Represents the geometric center of the point cloud composed of the remaining sampling points after denoising, Represents the total number of remaining sampling points after denoising, is the coordinate of the i-th sampling point in the point cloud composed of the remaining sampling points after denoising; S232, translate the point cloud to a position with the geometric center as the origin, and update the coordinate data of the point cloud, as shown in the following formula (5): (5); In formula (5), Represents the coordinates of the sampling point after translation, Represents the sampling coordinates before translation ; S233, calculate the maximum distance from all denoised sampling points to the origin after translation , see the following formula (6) for details: (6); S234. Scale the point cloud to the unit sphere so that the maximum distance between all points in the point cloud is 1. See the following formula (7): (7); In formula (7), Represents the normalized point coordinates; (3) Use the grid-based point cloud segmentation algorithm to separate the point cloud data of the wall prefabricated components from the point cloud image. The specific steps are as follows: S31. Define the grid size and then perform grid mapping, as shown in the following formula (8): (8); In formula (8), is the coordinate of any point in the preprocessed point cloud data after grid mapping, is the coordinate of any point in the preprocessed point cloud data, 、 、 is the three-dimensional size of the grid; S32. Calculate the geometric features of the point cloud in each grid cell, including the normal vector and curvature. The calculation of the normal vector is shown in the following formula (9): (9); In formula (9), represents the normal vector of the i-th grid cell, represents the coordinates of the points within the grid cell, represents the coordinates of the center point of the grid cell, Represents the number of points within a grid cell; The curvature calculation first calculates the local covariance matrix of the i-th grid cell , see the following formula (10) for details: (10); In formula (10), Represents the transpose operation; Then calculate the covariance matrix The eigenvalue of 、 and ,and , and then calculate the curvature according to formula (11) ; (11); S33. Calculate the data density of the point cloud in each grid cell, as shown in the following formula (12): (12); In formula (12), represents the point cloud density of the i-th grid cell, represents the volume of the i-th grid cell; S34. Setting threshold and , the data density The grid cells within the threshold range are divided into wall point clouds, as shown in the following formula (13): (13); In formula (13), Represents the wall point cloud; S35. Segmentation based on geometric features: further refine the segmentation results based on the normal vector and curvature geometric features to remove the mis-segmented point clouds in the background. The specific steps are as follows: Based on formula (9), the normal vectors of all grid cells are calculated and screened according to the direction consistency of the normal vectors. The grid cells with inconsistent normal vector directions are screened and removed. Calculate the unit normal vector of any two points within a single grid cell in the remaining grid cells and The angle between , see the following formula (14) for details: (14); In formula (14), Represents the modulus of the vector; and is the unit normal vector of any two points within a single grid cell. The steps for calculating the unit normal vector of a point within a single grid cell are: First, select any point within a single grid cell , construct its neighborhood point set : ; Calculate the neighborhood point set Center point , see the following formula (15) for details: (15); In formula (15), Representative Points Neighborhood points of Represents a set of neighborhood points The total number of neighborhood points in , ; Then the center point , neighborhood point set Substitute all the neighboring points in into formula (10) to calculate the neighboring point set The local covariance matrix of Replace the formula (10) , Replace the formula (10) , Replace the formula (10) ; Then calculate the neighborhood point set The eigenvalues ​​of the local covariance matrix 、 and ,and , 、 and Corresponding to three eigenvectors 、 and , the smallest eigenvalue The corresponding eigenvector That is the point The normal vector of point , is normalized according to the following formula (16) to obtain The unit normal vector ; (16); In formula (16), Yes The corresponding unit normal vector, Represents the vector modulus; Set the maximum normal vector angle of the precast wall component structure surface , when the angle between all normal vectors in the grid cell is No more than , it is preliminarily determined that the grid unit belongs to the structural surface of the prefabricated wall component, otherwise the point cloud data in the grid unit is filtered and removed; Then perform curvature judgment and set the curvature threshold , when the ith grid cell calculated by equation (11) No more than , it is finally determined that the grid unit belongs to the structural surface of the prefabricated wall component, that is, the point cloud data of the prefabricated wall component is separated from the point cloud image; otherwise, the point cloud data in the grid unit is filtered and removed, and the point cloud data in the remaining grid units are the point cloud data of the prefabricated wall component; (4) Construct a PointNet model to extract the feature points of the boundary of the prefabricated wall components, and obtain the boundary points of the prefabricated wall components through the DBSCAN density clustering algorithm. Calculate the length and width of the prefabricated wall components based on the coordinate information of the boundary points. The specific steps are as follows: S41. Build a PointNet model. The PointNet model includes an input layer, a feature extraction layer, a global feature layer, and an output layer. The input layer is used to input the point cloud data of prefabricated wall components. The three-dimensional coordinate information of each point is ; The feature extraction layer extracts features from each point through multiple fully connected layers and ReLU activation functions. The initial feature of the mth point in the wall prefabricated component point cloud data is , after being processed by the fully connected layer and ReLU activation function of the jth layer, the new feature is obtained ,in and are the weights and biases of the jth layer in the feature extraction layer; The global feature layer aggregates the features of all points through the maximum pooling operation to obtain the global feature , Represents the local features of each point, is the index of the last layer of feature extraction layer; The output layer is used to convert the global features After passing through the fully connected layer, the final result is output, that is, the global features are first Broadcast to each point, and then combine with the local features of each point Splicing to form the comprehensive features of each point , Represents splicing, and then the probability of each point belonging to a boundary point is calculated through the fully connected layer and sigmoid activation function , see the following formula (17) for details; (17); In formula (17), and are the weights and biases of the fully connected layers in the output layer, Represents the sigmoid activation function; S42. Set a probability threshold ,Will Greater than The points are identified as feature points of the boundary of prefabricated wall components; S43, using the DBSCAN density clustering algorithm to obtain the clustering clusters of the boundary part of the wall prefabricated components, first set the neighborhood radius And the minimum number of points minPts, then the feature points of the identified wall prefabricated component boundary are input into the DBSCAN density clustering algorithm, and the feature point set of the wall prefabricated component boundary is , feature points The coordinates are ,The DDBSCAN density clustering algorithm is used to divide the feature points of the boundary of the wall prefabricated components into multiple clusters, each cluster represents a boundary part of the wall; S44. The feature point set of a boundary part of the wall is , feature points The coordinates are , calculate the centroid coordinates of each boundary segment , see the following formula (18) for details: (18); In formula (18), Representative point set Total number of midpoints; S45. Calculate the length and width of the wall: The centroids of the left and right boundary parts of the wall are and ,but and The coordinates of and ; The centroids of the upper and lower boundary parts of the wall are and ,but and The coordinates of and ; then the length of the wall and width Calculated by the following formula (19) and formula (20) respectively; (19); (20); (5) In the point cloud image of the prefabricated wall component, use a visualization tool to visualize the point cloud and display the calculated length and width data of the prefabricated wall component on the point cloud image. The specific steps are as follows: S51, firstly mark the dimension data, and mark the length and width data of the detected prefabricated wall components on the point cloud visualization interface; S52. Draw dimension lines on the point cloud visualization interface to display the measured dimension range; the dimension lines consist of two parallel lines and a vertical line connecting the two parallel lines, and the dimension values ​​are marked on the vertical line; and perform text annotation, add text labels on the point cloud visualization interface to display the specific values ​​of the dimension data, specifically, mark "length" and "width" on the upper edge and left edge of the prefabricated wall component respectively.

[0021] Performance Analysis: The experiment collected a total of 304 wall datasets, including 246 training datasets and 58 test datasets. These datasets encompass prefabricated wall panels of varying sizes and specifications. In this experiment, a 304-sample wall dataset was collected to cover prefabricated wall panels of varying sizes and specifications, ensuring data diversity and representativeness. To fully utilize the data and reasonably evaluate model performance, the dataset was divided into a training set and a test set. The training set contained 246 samples, approximately 81% of the total, and was used for model training and optimization. The test set contained 58 samples, approximately 19% of the total, and was used for model testing and evaluation.

[0022] The detection method in this embodiment of the present invention was compared with the traditional Candy operator detection method. The comparative experimental results are shown in Table 1 below. In this experiment, this embodiment of the present invention combined the PointNet model with the BSCAN density clustering algorithm to extract and calculate the length and width dimensions of prefabricated components. The traditional Candy operator detection method has certain limitations in extracting dimensional information, especially when processing complex shapes and high-density point cloud data, where its efficiency and accuracy are relatively poor.

[0023] This embodiment's detection method, based on the PointNet model and the BSCAN density clustering algorithm, achieves significant improvements in both average detection time and average detection error, resulting in an approximately 30% improvement in overall detection performance compared to the Candy operator detection method. This improvement not only enhances the efficiency and accuracy of prefabricated component dimensional inspection but also provides stronger technical support for prefabricated component quality control and production management, demonstrating significant practical application value.

[0024] Table 1

[0025] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting the size of prefabricated walls using point clouds based on deep learning, characterized by: The specific steps include: (1) Collect point cloud data of the structural surface of the prefabricated wall components to be tested and construct a point cloud image dataset; (2) Preprocess the point cloud data in the point cloud image to remove noise points and redundant data points to obtain preprocessed point cloud data; (3) Use the grid-based point cloud segmentation algorithm to separate the point cloud data of the wall prefabricated components from the point cloud image; (4) Construct a PointNet model to extract the feature points of the boundary of the prefabricated wall components, and obtain the clusters of the boundary of the prefabricated wall components through the DBSCAN density clustering algorithm. Calculate the length and width of the prefabricated wall components based on the cluster information; (5) In the point cloud image of the prefabricated wall component, a visualization tool is used to visualize the point cloud, and the calculated length and width data of the prefabricated wall component are displayed on the point cloud image.

2. The method for detecting prefabricated wall dimensions based on point cloud based on deep learning according to claim 1, characterized in that: The point cloud data of the structural surface of the prefabricated wall component to be measured is collected by using a 3D laser camera.

3. The method for detecting prefabricated wall dimensions based on point cloud based on deep learning according to claim 1, characterized in that: The specific steps of preprocessing the point cloud data are as follows: S21, downsampling processing: select the set radius The point cloud space is divided into two parts. The point closest to the center of each sphere is selected as the sampling point to replace all the points in the sphere. Only the sampling point of each sphere is retained in the point cloud, and the position of the sampling point does not move. S22, denoising: for each of the remaining sampling points, search its neighbor point set within a given radius. When the number of neighbor points in the neighbor point set is less than the set minimum neighbor number threshold, mark the point as an outlier and remove it from the sampling points. S23, normalization processing: normalize the remaining sampling points after denoising to eliminate the scale difference of the point cloud data so that point cloud data of different sizes have the same scale.

4. The method for detecting prefabricated wall dimensions based on point cloud based on deep learning according to claim 3, characterized in that: The specific steps of the downsampling process are: S211, select the center point of the sphere, randomly select an initial point from the point cloud as the center point of the sphere , and then search and confirm the points in the sphere, that is, calculate all the points in the point cloud and the center point The distance is not greater than the set radius The points are included in the current sphere, as shown in the following formula (1): （1）； In formula (1), Represents the set of points within the current sphere, Represents a point within the current sphere, Represents a point in the point cloud, represents the center point of the sphere, represents the Euclidean distance, Represents the set radius; S212. The set of points inside the current sphere , find the distance from the center point The nearest point , as the sampling point of the current sphere, use the sampling point to replace other points in the current sphere, the position of the sampling point does not move, the sampling point The selection process is shown in the following formula (2): （2）； In formula (2), Represents any point in the current sphere except the center point. represents the center point of the sphere, represents the Euclidean distance; S213. Update the point cloud and sphere center: remove all points in the searched sphere, randomly select a new sphere center point from the remaining points in the point cloud, repeat steps S21 and S22 to search for points in the sphere, and update the sampling points of the sphere until all points in the point cloud data are searched and processed or the required number of sampling points is reached.

5. The method for detecting prefabricated wall dimensions based on point cloud based on deep learning according to claim 3, characterized in that: The specific steps of the denoising process are: S221, for each sampling point , calculate its The set of neighbor points in , see the following formula (3) for details: （3）； In formula (3), Representative sampling points and sampling points The Euclidean distance between S222, when the sampling point The set of neighbor points The number of neighbor points in the network is less than the set minimum number of neighbors. When , the sampling point is removed from the sampling point set .

6. The method for detecting prefabricated wall dimensions based on point cloud based on deep learning according to claim 3, characterized in that: The specific steps of the normalization process are: S231. Calculate the geometric center of the point cloud, as shown in the following formula (4): （4）； In formula (4), Represents the geometric center of the point cloud composed of the remaining sampling points after denoising, Represents the total number of remaining sampling points after denoising, is the coordinate of the i-th sampling point in the point cloud composed of the remaining sampling points after denoising; S232, translate the point cloud to a position with the geometric center as the origin, and update the coordinate data of the point cloud, as shown in the following formula (5): （5）； In formula (5), Represents the coordinates of the sampling point after translation, Represents the sampling coordinates before translation ; S233, calculate the maximum distance from all denoised sampling points to the origin after translation , see the following formula (6) for details: （6）； S234. Scale the point cloud to the unit sphere so that the maximum distance between all points in the point cloud is 1. See the following formula (7): （7）； In formula (7), Represents the normalized point coordinates.

7. The method for detecting prefabricated wall dimensions based on point cloud based on deep learning according to claim 3, characterized in that: The specific steps of using the grid-based point cloud segmentation algorithm to separate the point cloud data of the prefabricated wall components from the point cloud image are as follows: S31. Define the grid size and then perform grid mapping, as shown in the following formula (8): （8）； In formula (8), is the coordinate of any point in the preprocessed point cloud data after grid mapping, is the coordinate of any point in the preprocessed point cloud data, 、 、 is the three-dimensional size of the grid; S32. Calculate the geometric features of the point cloud in each grid cell, including the normal vector and curvature, to provide the necessary geometric information for subsequent segmentation. The directional consistency of the normal vector is the key basis for determining whether the point clouds belong to the same area. The calculation of the normal vector is shown in the following formula (9): （9）； In formula (9), represents the normal vector of the i-th grid cell, represents the coordinates of the points within the grid cell, represents the coordinates of the center point of the grid cell, Represents the number of points within a grid cell; The curvature calculation first calculates the local covariance matrix of the i-th grid cell , see the following formula (10) for details: （10）； In formula (10), Represents the transpose operation; Then calculate the covariance matrix The eigenvalue of 、 and ,and , and then calculate the curvature according to formula (11) ; （11）； S33. Calculate the data density of the point cloud in each grid cell, as shown in the following formula (12): （12）； In formula (12), represents the point cloud density of the i-th grid cell, represents the volume of the i-th grid cell; S34. Setting threshold and , the data density The grid cells within the threshold range are divided into wall point clouds, as shown in the following formula (13): （13）； In formula (13), Represents the wall point cloud; S35. Segmentation based on geometric features: further refine the segmentation results based on the normal vector and curvature geometric features to remove the mis-segmented point clouds in the background. The specific steps are as follows: Based on formula (9), the normal vectors of all grid cells are calculated and screened according to the direction consistency of the normal vectors. The grid cells with inconsistent normal vector directions are screened and removed. Calculate the unit normal vector of any two points within a single grid cell in the remaining grid cells and The angle between , see the following formula (14) for details: （14）； In formula (14), and is the unit normal vector between any two points within a single grid cell, Represents the modulus of the vector; Set the maximum normal vector angle of the precast wall component structure surface , when the angle between all normal vectors in the grid cell is No more than , it is preliminarily determined that the grid unit belongs to the structural surface of the prefabricated wall component, otherwise the point cloud data in the grid unit is filtered and removed; Then perform curvature judgment and set the curvature threshold , when the ith grid cell calculated by equation (11) No more than , it is finally determined that the grid unit belongs to the structural surface of the prefabricated wall component, that is, the point cloud data of the prefabricated wall component is separated from the point cloud image; otherwise, the point cloud data in the grid unit is filtered and removed, and the point cloud data in the remaining grid units are the point cloud data of the prefabricated wall component.

8. The method for detecting prefabricated wall dimensions based on point cloud based on deep learning according to claim 7, characterized in that: The steps for calculating the unit normal vector of a point within a single grid cell are: First, select any point within a single grid cell , construct its neighborhood point set : ； Calculate the neighborhood point set Center point , see the following formula (15) for details: （15）； In formula (15), Representative Points The neighborhood points are the distance points in the point cloud data nearest point; Table neighborhood point set The total number of neighborhood points in the Then the center point , neighborhood point set Substitute all the neighboring points in into formula (10) to calculate the neighboring point set The local covariance matrix of Replace the formula (10) , Replace the formula (10) , Replace the formula (10) ; Then calculate the neighborhood point set The eigenvalues ​​of the local covariance matrix 、 and ,and , 、 and Corresponding to three eigenvectors 、 and , the smallest eigenvalue The corresponding eigenvector That is the point The normal vector of point , is normalized according to the following formula (16) to obtain The unit normal vector ; （16）； In formula (16), Yes The corresponding unit normal vector, Represents the vector modulus.

9. The method for detecting prefabricated wall dimensions based on point cloud based on deep learning according to claim 7, characterized in that: The specific steps of constructing the PointNet model, extracting the feature points of the boundary of the prefabricated wall components, and obtaining the clusters of the boundary of the prefabricated wall components through the DBSCAN density clustering algorithm, and calculating the length and width of the prefabricated wall components based on the cluster information are as follows: S41. Build a PointNet model. The PointNet model includes an input layer, a feature extraction layer, a global feature layer, and an output layer. The input layer is used to input the point cloud data of prefabricated wall components. The three-dimensional coordinate information of each point is ; The feature extraction layer extracts features from each point through multiple fully connected layers and ReLU activation functions. The initial feature of the mth point in the wall prefabricated component point cloud data is , after being processed by the fully connected layer and ReLU activation function of the jth layer, the new feature is obtained ,in and are the weights and biases of the jth layer in the feature extraction layer; The global feature layer aggregates the features of all points through the maximum pooling operation to obtain the global feature , Represents the local features of each point, is the index of the last layer of feature extraction layer; The output layer is used to convert the global features After passing through the fully connected layer, the final result is output, that is, the global features are first Broadcast to each point, and then combine with the local features of each point Splicing to form the comprehensive features of each point , Represents splicing, and then the probability of each point belonging to a boundary point is calculated through the fully connected layer and sigmoid activation function , see the following formula (17) for details; （17）； In formula (17), and are the weights and biases of the fully connected layers in the output layer, Represents the sigmoid activation function; S42. Set a probability threshold ,Will Greater than The points are identified as feature points of the boundary of prefabricated wall components; S43, using the DBSCAN density clustering algorithm to obtain the clustering clusters of the boundary part of the wall prefabricated components, first set the neighborhood radius And the minimum number of points minPts, then the feature points of the identified wall prefabricated component boundary are input into the DBSCAN density clustering algorithm, and the feature point set of the wall prefabricated component boundary is , feature points The coordinates are ,The DDBSCAN density clustering algorithm is used to divide the feature points of the boundary of the wall prefabricated components into multiple clusters, each cluster represents a boundary part of the wall; S44. The feature point set of a boundary part of the wall is , feature points The coordinates are , calculate the centroid coordinates of each boundary segment , see the following formula (18) for details: （18）； In formula (18), Representative point set Total number of midpoints; S45. Calculate the length and width of the wall: The centroids of the left and right boundary parts of the wall are and ,but and The coordinates of and The centroids of the upper and lower boundary parts of the wall are and ,but and The coordinates of and ; then the length of the wall and width Calculated by the following formula (19) and formula (20) respectively; （19）； （20）。

Images