Multi-line structured light three-dimensional reconstruction frame point cloud real-time de-stacking and fusion method

Abstract

The invention provides a multi-line structured light three-dimensional reconstruction frame point cloud real-time de-stacking and fusion method, and relates to the technical field of line structured light three-dimensional reconstruction. The method comprises the following steps: establishing a global grid structure through the position of a multi-view frame point cloud during three-dimensional reconstruction by utilizing the distribution characteristics of the spatial position of the point cloud, solving a field point set corresponding to each frame of point cloud, and performing normal feature estimation according to the distribution characteristics of neighborhood points; and establishing a cylindrical neighborhood structure for the overlapped frame point clouds according to the overlapping condition of the point clouds of the global grid structure, and realizing point cloud fusion through vector direction weighting and normal vector features of the point clouds of the target point and the fusion neighborhood point set. According to the algorithm, the system stability requirement in three-dimensional reconstruction can be met, the quality of a final point cloud model is improved, the complex environment measurement adaptability of multi-line structured light three-dimensional reconstruction is greatly improved, and the multi-line structured light three-dimensional reconstruction precision is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of line structured light three-dimensional reconstruction, and in particular to a real-time de-stacking and fusion method for multi-line structured light three-dimensional reconstruction frame point clouds. Background Art

[0002] The structured light scanning system based on the principle of stereoscopic vision is one of the effective methods to obtain the three-dimensional surface data of objects. It is widely used in reverse engineering, biomedicine, cultural relics protection and other fields, and is an important way to achieve workpiece size measurement and defect detection. Like other three-dimensional scanning devices, multi-line structured light scanners can quickly reconstruct the surface of objects, but due to their limited scanning field of view, images of objects need to be collected from different perspectives to reconstruct a complete three-dimensional surface model of the object. Therefore, during the scanning process, the point clouds reconstructed from different perspectives need to be registered with the point clouds in the global coordinate system. In order to make the reconstructed model more complete, a certain position will be scanned multiple times during reconstruction, and the final laser point cloud model will inevitably have problems such as stacking, redundancy, and uneven distribution. In addition, when scanning and measuring large-sized objects, the registration accuracy of multi-view point cloud data will decrease with the increase in the number of registrations. Due to insufficient registration accuracy, the point cloud model will have local delamination lines. If the point cloud is directly constructed into a patch model, there will be more redundant patches on the model surface, the surface will be too rough, and the splicing marks will be obvious. In order to avoid the influence of multiple scans and stitching of the same location, and obtain a uniform and smooth point cloud model, the point cloud model needs to be fused. Point cloud fusion is the basis of 3D reconstruction. It can not only fuse multiple layers of point cloud data into a single layer of point cloud data to reduce point cloud data redundancy, but also remove noise interference to obtain a model with unchanged features and smooth surface.

[0003] Point cloud fusion can be divided into multi-view frame point cloud fusion processing and overall point cloud model post-processing in terms of processing methods. Overall point cloud model post-processing is to fuse the entire point cloud model to obtain a uniform, single-layer point cloud model after the scanning and reconstruction of the entire measured object is completed. The post-processing method is difficult to consider the surface point cloud deviation during the reconstruction process, and it is difficult to process point clouds with uneven surface thickness distribution, and it is impossible to effectively and accurately compensate the point cloud surface. Summary of the invention

[0004] In order to overcome the deficiencies of the prior art, an object of the present invention is to provide a method for real-time de-stacking and fusion of multi-line structured light three-dimensional reconstruction frame point clouds.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A method for real-time de-stacking and fusion of multi-line structured light 3D reconstruction frame point cloud, comprising:

[0007] Acquire laser point cloud data of the target to be reconstructed and perform preprocessing to obtain preprocessed laser point cloud data;

[0008] Establishing an initial network structure according to the three-dimensional coordinates of the first frame point cloud in the laser point cloud data;

[0009] Using subsequent frame point clouds in the laser point cloud data to classify point clouds and expand the network structure on the initial network structure to obtain a final network structure, and determining the index of the common grid and the frame point cloud and the total point cloud in the common grid according to the final network structure;

[0010] Establishing a corresponding domain network structure according to the target point in the public network;

[0011] Based on the corresponding domain network structure, searching for the central grid where the target point is located and its eight neighborhood grid point clouds in the x, y, and z axis directions to form a neighborhood point set of the target point;

[0012] According to the neighborhood point set and the principal component analysis method, the normal direction of the target point is estimated and the normal bidirectionality is eliminated to obtain an optimized normal vector;

[0013] Determine redundant points of the target point according to the optimized normal vector and remove the redundant points to obtain laser point cloud data after removal;

[0014] Establishing a cylindrical neighborhood of the target point according to the removed laser point cloud data;

[0015] The target points are optimized and iterated according to the cylindrical domain to obtain a fused point cloud.

[0016] Preferably, the step of establishing an initial network structure according to the three-dimensional coordinates of the first frame point cloud in the laser point cloud data includes:

[0017] According to the three-dimensional coordinates of the first frame point cloud, the maximum coordinates and the minimum coordinates of the first frame point cloud are calculated to obtain a spatial bounding box of the first frame point cloud;

[0018] Calculate the average point spacing of the point cloud;

[0019] Calculate the grid side length according to the average point spacing;

[0020] The initial network structure is obtained by numbering the center coordinates according to the side length of the telescopic grid and the spatial bounding box.

[0021] Preferably, the method of using subsequent frame point clouds in the laser point cloud data to classify point clouds and expand the network structure on the initial network structure to obtain a final network structure and determining a common grid according to the final network structure includes:

[0022] Calculate the Euclidean minimum distance from the subsequent frame point cloud to the center point of the initial network structure;

[0023] According to the Euclidean minimum distance of the center point, it is judged whether the subsequent frame point cloud exceeds the range of the initial network structure to obtain a judgment result. If the judgment result is yes, the current corresponding point cloud is stored. If the judgment result is no, the network structure of the current point cloud is expanded to obtain a public network.

[0024] Preferably, the step of estimating the normal direction of the target point and eliminating the bidirectionality of the normal to obtain an optimized normal vector according to the neighborhood point set and the principal component analysis method includes:

[0025] Calculate the centroid of the preset point in the field point set according to the target point;

[0026] Constructing a corresponding covariance matrix according to the centroid and the preset points in the domain point set;

[0027] Decomposing the covariance matrix using singular values ​​to obtain eigenvalues ​​and eigenvectors;

[0028] Obtaining a normal vector of the target point according to the eigenvalue and the eigenvector;

[0029] The normal vector of the target point is bidirectionally eliminated to obtain an optimized normal vector.

[0030] Preferably, the step of determining redundant points of the target point according to the optimized normal vector and removing the redundant points to obtain laser point cloud data after removal includes:

[0031] The corresponding direction vector is calculated based on the optimized normal vector and the normal vector corresponding to the preset point in the domain point set;

[0032] Calculate the angle between the direction vector and the optimized normal vector and the normal vector corresponding to the preset point to obtain a first angle and a second angle;

[0033] Redundant points are determined according to the angle difference between the first angle and the second angle, and are eliminated to obtain laser point cloud data after elimination.

[0034] Preferably, the step of establishing a cylindrical neighborhood of the target point based on the removed laser point cloud data comprises:

[0035] Using the redundant point set corresponding to the domain point set to correct the normal vector of the target point in the laser point cloud data after the elimination, to obtain a corrected normal vector;

[0036] Calculate the projection points of the target point along the approximate corrected normal vector in the positive and negative directions to determine the center coordinates of the upper and lower surfaces of the cylinder;

[0037] The cylinder neighborhood is determined according to preset cylinder parameters and the center coordinates.

[0038] Preferably, the optimizing and iterating the target points according to the cylindrical domain to obtain a fused point cloud comprises:

[0039] Identify redundant reference points;

[0040] Optimizing and iterating the target point according to the redundant reference point to obtain a fusion point position;

[0041] The fused point cloud is obtained according to the fused point positions.

[0042] Preferably, the expression of the redundant reference point is:

[0043]

[0044] Among them, p i is the target point, p j is the neighborhood point, v ij is the direction vector, d i With d j They are v ij In p i and p j The forward projection length in the normal direction, p c Redundant reference point.

[0045] The present invention discloses the following technical effects:

[0046] The present invention provides a real-time de-stacking and fusion method for multi-line structured light 3D reconstruction frame point cloud, comprising: obtaining laser point cloud data of a target to be reconstructed and preprocessing it to obtain preprocessed laser point cloud data; establishing an initial network structure according to the 3D coordinates of the first frame point cloud in the laser point cloud data; using subsequent frame point clouds in the laser point cloud data to perform point cloud classification and network structure expansion on the initial network structure to obtain a final network structure and determining a common grid and an index of a frame point cloud and a total point cloud in the common grid according to the final network structure; establishing a corresponding domain network structure according to target points in the common network ; Based on the corresponding domain network structure, search the central grid where the target point is located and its eight neighborhood grid point clouds in the x, y, and z axis directions to form a neighborhood point set of the target point; estimate the normal of the target point based on the neighborhood point set and principal component analysis method and eliminate the normal bidirectionality to obtain an optimized normal vector; determine the redundant points of the target point based on the optimized normal vector and eliminate the redundant points to obtain the eliminated laser point cloud data; establish the cylindrical neighborhood of the target point based on the eliminated laser point cloud data; optimize and iterate the target point based on the cylindrical domain to obtain a fused point cloud. According to the spatial distribution characteristics of the point cloud, the present invention establishes a global grid structure and a cylindrical fusion domain, and proposes a multi-view frame point cloud real-time fusion algorithm for multi-line structured light 3D reconstruction based on normal constraints and point cloud direction vector weighting. The algorithm realizes denoising and classification preprocessing of multi-line structured light point clouds, establishes a spatial global grid structure based on the spatial bounding box of the first frame point cloud, and classifies the point cloud grid and expands the grid structure on the spatial structure of the first frame point cloud based on the Euclidean distance of the point cloud for subsequent frame point clouds. According to the global grid structure, the relative position of each frame point cloud in the cumulatively reconstructed global point cloud is obtained, and the points in the grid of the single frame point cloud that are classified into the existing point cloud are fused. During the fusion process, a 3×3×3 grid structure is established, and the current frame point cloud and the cumulatively reconstructed global point cloud in the common grid are obtained, and the obtained global point cloud is used as the feature estimation neighborhood point set of the point to be fused. Finally, a spatial cylindrical structure of the point to be fused is established in the field point set, and the point cloud normal fusion based on the guidance of the point cloud direction vector is realized in the cylindrical neighborhood. The present invention effectively reduces the resource consumption of the point cloud fusion algorithm and improves the fusion efficiency. It can make the point cloud of the reconstructed object point cloud model evenly distributed, and has a good adjustment effect on the overall point cloud density. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0048] Figure 1 It is a flow chart of a method for real-time de-stacking and fusion of multi-line structured light 3D reconstruction frame point cloud of the present invention;

[0049] Figure 2 It is a schematic diagram of the central grid 3×3×3 structure neighborhood point cloud search of the present invention;

[0050] Figure 3 A schematic diagram of determining a redundant point set of a target point and establishing a neighborhood of a cylinder according to the present invention;

[0051] Figure 4 This is a schematic diagram of the point cloud effect before and after the fusion of the cultural relics scanning and reconstruction of the present invention;

[0052] Figure 5 It is a schematic diagram of the effect of fusion of point clouds before and after the cylindrical scanning reconstruction of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0055] like Figure 1 As shown, the present invention provides a method for real-time de-stacking and fusion of multi-line structured light 3D reconstruction frame point cloud, comprising:

[0056] Step 100: Acquire laser point cloud data of the target to be reconstructed and perform preprocessing to obtain preprocessed laser point cloud data;

[0057] Specifically, a binocular multi-line structured light 3D reconstruction system is used to reconstruct and acquire laser data, and the multi-line structured light point cloud is denoised and classified preprocessed:

[0058] Point cloud preprocessing includes denoising and classification preprocessing of the multi-line structured light point cloud obtained by scanning; in view of the irregular arrangement of the extracted laser line point cloud and the existence of a small number of isolated points and low-point-count point cloud clusters in a single laser line, the Euclidean clustering algorithm is used to remove point clouds with too low point counts in the point cloud clusters and retain point clouds with higher continuity.

[0059] Step 200: establishing an initial network structure according to the three-dimensional coordinates of the first frame point cloud in the laser point cloud data;

[0060] Step 300: using subsequent frame point clouds in the laser point cloud data to classify point clouds and expand the network structure on the initial network structure to obtain a final network structure, and determining the index of the common grid and the frame point cloud and the total point cloud in the common grid according to the final network structure;

[0061] Step 400: Establishing a corresponding domain network structure according to the target point in the public network;

[0062] Step 500: Based on the corresponding domain network structure, search for the central grid where the target point is located and its eight neighborhood grid point clouds in the x, y, and z axis directions to form a neighborhood point set of the target point;

[0063] Step 600: estimating the normal direction of the target point according to the neighborhood point set and the principal component analysis method and eliminating the bidirectionality of the normal to obtain an optimized normal vector;

[0064] Step 700: determining redundant points of the target point according to the optimized normal vector and eliminating the redundant points to obtain laser point cloud data after elimination;

[0065] Step 800: establishing a cylindrical neighborhood of the target point according to the laser point cloud data after elimination;

[0066] Step 900: Optimize and iterate the target point according to the cylindrical domain to obtain a fused point cloud.

[0067] Further, such as Figure 2 As shown, the initial network structure is established according to the three-dimensional coordinates of the first frame point cloud in the laser point cloud data, including:

[0068] According to the three-dimensional coordinates of the first frame point cloud, the maximum coordinates and the minimum coordinates of the first frame point cloud are calculated to obtain a spatial bounding box of the first frame point cloud;

[0069] Calculate the average point spacing of the point cloud;

[0070] Calculate the grid side length according to the average point spacing;

[0071] The initial network structure is obtained by numbering the center coordinates according to the side length of the telescopic grid and the spatial bounding box.

[0072] Specifically, the bounding box is solved according to the spatial position of the point cloud of the first frame, and a spatial global grid structure is established;

[0073] Subsequent frame point clouds are classified into point cloud grids and grid structure expanded based on the spatial structure of the first frame point cloud using the Euclidean distance of the point clouds.

[0074] The method of establishing a global grid structure for the first frame point cloud is as follows:

[0075] According to the three-dimensional coordinates of the point cloud in the first frame, the maximum and minimum coordinates of the point cloud are calculated to obtain the spatial bounding box of the point cloud:

[0076] V={(x min ,y min ,y min ),(x max ,y max ,y max )};

[0077] According to the average distance between adjacent points in the point cloud, the grid side length, the total number of point cloud points, coordinates and other parameters are set to calculate the average distance d avg , take d avg The integer multiple k of is used as the grid side length;

[0078] According to the determined point cloud side length L, the minimum bounding box of the point cloud in the x-axis, y-axis, and z-axis directions in space is numbered V (v i ,v j ,v k ).

[0079] Furthermore, the method of using subsequent frame point clouds in the laser point cloud data to classify point clouds and expand the network structure on the initial network structure to obtain a final network structure and determining a common grid according to the final network structure includes:

[0080] Calculate the Euclidean minimum distance from the subsequent frame point cloud to the center point of the initial network structure;

[0081] According to the Euclidean minimum distance of the center point, it is judged whether the subsequent frame point cloud exceeds the range of the initial network structure to obtain a judgment result. If the judgment result is yes, the current corresponding point cloud is stored. If the judgment result is no, the network structure of the current point cloud is expanded to obtain a public network.

[0082] Specifically, determine each 3D point p in the current frame point cloud i (x i ,y i ,z i ) and the global grid structure center coordinates (v i ,v j ,v k ) is the minimum Euclidean distance d i , the distance within the frame point cloud satisfies d i Points with coordinates less than L are classified into the grid center (v i ,v j ,v k ) in the grid index;

[0083] Traverse all existing point cloud grids, expand and store the grid structure for points that do not meet all grid center distance constraints; set the spatial bounding box coordinates of the first frame to be The spatial bounding box coordinates of subsequent frames are In order to ensure the continuity and unity of the spatial grid structure, the minimum coordinates of the bounding box of the first frame are used. The starting point of the space grid.

[0084] Furthermore, the method of estimating the normal direction of the target point and eliminating the bidirectionality of the normal line to obtain an optimized normal vector according to the neighborhood point set and the principal component analysis method includes:

[0085] Calculate the centroid of the preset point in the field point set according to the target point;

[0086] Constructing a corresponding covariance matrix according to the centroid and the preset points in the domain point set;

[0087] Decomposing the covariance matrix using singular values ​​to obtain eigenvalues ​​and eigenvectors;

[0088] Obtaining a normal vector of the target point according to the eigenvalue and the eigenvector;

[0089] The normal vector of the target point is bidirectionally eliminated to obtain an optimized normal vector.

[0090] Specifically, the relative position of each frame point cloud in the cumulatively reconstructed global point cloud is obtained according to the global grid structure, and the grid with overlapping point clouds and the neighborhood grid point cloud of the overlapping grid are obtained;

[0091] The normal vector of the neighborhood point cloud is estimated according to the area point set where the target point of the single-frame point cloud is located, and the bidirectionality of the normal vector is eliminated.

[0092] The method for obtaining the relative position of the current frame point cloud in the cumulatively reconstructed global point cloud and the neighborhood point cloud is as follows: after the current grid is added with a certain frame of point cloud data, if the subsequent frame point cloud is also merged into the grid, the grid may have redundant point clouds and is regarded as the area to be fused; the grid of the area to be fused is taken as the center, and a 3×3×3 structure of the central grid is established; starting from the central grid, eight neighborhood grids are searched layer by layer in the positive and negative directions of the x-axis, y-axis, and z-axis to form the nearest voxel grid structure, and the total point cloud in the 3×3×3 grid is taken as the target point p i (x i ,y i ,z i )’s neighborhood normal feature estimation point cloud set

[0093] Mark the global grid structure and the center grid neighborhood search results. The point cloud set of the i-th frame is The point cloud set in the frame point cloud that belongs to the common grid of the existing point cloud is The newly added point cloud set in the frame point cloud is

[0094] The total point cloud set accumulated by 3D reconstruction is S global , determine the target point p of the frame point cloud in the public grid during neighborhood search i (x i ,y i ,z i ) corresponds to the total point cloud point set in the same grid And according to the central grid 3×3×3 structure Point cloud collection, and satisfy the relationship

[0095] The specific method for estimating the normal feature of the point cloud in the area where the target point is located and eliminating the bidirectionality of the normal vector is as follows:

[0096] The principal component analysis (PCA) is used to estimate the target point p i (x i ,y i ,z i ) is the neighborhood point set The normal vector characteristics of Normals of each point in the i (x i ,y i ,z i ) as the center, calculate the point p in its k neighborhood j (x j ,y j ,z j ) Then, the covariance matrix M in the k-neighborhood is constructed based on the neighborhood points and the centroid point, and the covariance matrix M is a symmetric matrix; finally, the singular value decomposition is used to obtain the eigenvalue λ of M r and the eigenvector v r , the minimum eigenvalue λ of the matrix M 0 The corresponding eigenvector v 0 That is p i The normal vector of

[0097]

[0098] right The normal direction obtained by the initial estimation is bidirectionally eliminated. In the process of multi-line structured light 3D reconstruction, the reconstructed object is always facing the scanning device. Each frame of laser point cloud extracted from the laser image is always along the optical center direction of the camera in the camera coordinate system; therefore, the angle β between the obtained laser point cloud coordinates and the surface point cloud normal vector of the measured object is only greater than 90°, less than 90°, and equal to 90°. For the target point p i (x i ,y i ,z i ) corresponds to the normal vector n i (x i ,y i ,z i ) for correction; specify the point cloud normal vector n i Satisfying β≥90°, when β<90°, let n i =-n i , thus achieving fast global redirection of normal vectors.

[0099] Furthermore, the step of determining redundant points of the target point according to the optimized normal vector and removing the redundant points to obtain laser point cloud data after removal includes:

[0100] The corresponding direction vector is calculated based on the optimized normal vector and the normal vector corresponding to the preset point in the domain point set;

[0101] Calculate the angle between the direction vector and the optimized normal vector and the normal vector corresponding to the preset point to obtain a first angle and a second angle;

[0102] Redundant points are determined according to the angle difference between the first angle and the second angle, and are eliminated to obtain laser point cloud data after elimination.

[0103] Specifically, the redundancy of the point cloud is estimated according to the normal vector of the point cloud, and the redundant points of the target point cloud are determined by using the direction vectors of the target point and the neighboring points in the space and the projection of the normal vectors of the neighboring points in the vertical direction; the cylindrical neighborhood structure of the target point is solved according to the approximate normal direction of the target point.

[0104] Further, such as Figure 3-5 As shown, the step of establishing a cylindrical neighborhood of the target point based on the removed laser point cloud data includes:

[0105] Using the redundant point set corresponding to the domain point set to correct the normal vector of the target point in the laser point cloud data after the elimination, to obtain a corrected normal vector;

[0106] Calculate the projection points of the target point along the approximate corrected normal vector in the positive and negative directions to determine the center coordinates of the upper and lower surfaces of the cylinder;

[0107] The cylinder neighborhood is determined according to preset cylinder parameters and the center coordinates.

[0108] The method for solving the approximate normal direction of the target point cloud based on the normal vector characteristics of the neighborhood point set and the weighted Euclidean distance of the point cloud is as follows:

[0109] The PCA algorithm is used to estimate the normal vector n of the target point i and the neighborhood point p j The normal vector n j ; Calculate each neighborhood point p j and p i The direction vector v j , respectively find v j With n i and n j The angles θ and α are considered as p i The redundancy points are determined as follows.

[0110]

[0111] The specific method for estimating the circle domain of redundant point sets based on the normal vector characteristics of the neighborhood point set is:

[0112] According to the point set in the redundant neighborhood of the point cloud, the target point p i The normal vector n i make corrections;

[0113] Calculate the projection points of the target point along the approximate normal in the positive and negative directions. First, normalize the normal direction vector of the target point, calculate the projection distance according to the height H of the cylinder, and obtain the center coordinates of the upper and lower surfaces of the cylinder.

[0114] Find the point cloud set based on the cylinder's creation parameters The set of points inside the cylinder The target point is located at the center of mass of the cylinder, and the circles on the upper and lower surfaces have the same normal vector as the target point. calculate The direction vector between each point and the target point Distance d from the set j , and d j Then find the angle θ between the direction vector and the normal vector of the target point. According to the set radius R and height H of the cylinder, when the length of the point cloud distance in the normal and tangential directions of the target point is less than R and H respectively, it is determined to be a point inside the cylinder. The determination formula is as follows.

[0115]

[0116] Furthermore, the optimizing and iterating the target point according to the cylindrical domain to obtain a fused point cloud includes:

[0117] Identify redundant reference points;

[0118] Optimizing and iterating the target point according to the redundant reference point to obtain a fusion point position;

[0119] The fused point cloud is obtained according to the fused point positions.

[0120] Specifically, the redundant reference points in the neighborhood are determined, the direction vectors of each point in the cylinder neighborhood point set and the target point are calculated, and the angle between the direction vector and the normal vector of the target point is obtained. The neighborhood point with the smallest angle is used as the redundant reference point of the target point in the corresponding neighborhood.

[0121] Sliding iterations are performed in the direction of the target point and the reference point, and the point with the smallest cumulative error of redundant point iterations is taken as the fusion point.

[0122] The method for determining redundant reference points in the neighborhood is specifically as follows:

[0123] Determine the nearest reference point of the target point in the redundant neighborhood and calculate the target point p i With neighboring point p j The direction vector v ij , find v ij In p i and p j The forward projection length d in the normal direction i With d j ,|d i -d j The neighborhood point with the smallest |time is the reference point p of the target point c ;

[0124]

[0125] Among them, p i is the target point, p j is the neighborhood point, v ij is the direction vector, d i With d j They are v ij In p i and p j The forward projection length in the normal direction, p c Redundant reference point.

[0126] Find the direction vector between the reference point and the target point, and slide the target point on the direction vector for iteration. When the sum of the vector angle differences between the redundant point set and the non-redundant points in the neighborhood is the smallest, it means that the current iteration position has similar spatial distribution characteristics to the neighborhood position. The position of this point is used as the new point after fusion.

[0127] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0128] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for real-time de-stacking and fusion of multi-line structured light 3D reconstruction frame point cloud, characterized in that: include: Acquire laser point cloud data of the target to be reconstructed and perform preprocessing to obtain preprocessed laser point cloud data; Establishing an initial network structure according to the three-dimensional coordinates of the first frame point cloud in the laser point cloud data; Using subsequent frame point clouds in the laser point cloud data to classify point clouds and expand the network structure on the initial network structure to obtain a final network structure, and determining the index of the common grid and the frame point cloud and the total point cloud in the common grid according to the final network structure; Establishing a corresponding domain network structure according to the target point in the public network; Based on the corresponding domain network structure, searching for the central grid where the target point is located and its eight neighborhood grid point clouds in the x, y, and z axis directions to form a neighborhood point set of the target point; According to the neighborhood point set and the principal component analysis method, the normal direction of the target point is estimated and the normal bidirectionality is eliminated to obtain an optimized normal vector; Determine redundant points of the target point according to the optimized normal vector and remove the redundant points to obtain laser point cloud data after removal; Establishing a cylindrical neighborhood of the target point according to the removed laser point cloud data; The target points are optimized and iterated according to the cylindrical domain to obtain a fused point cloud.

2. The method for real-time de-stacking and fusion of multi-line structured light 3D reconstruction frame point cloud according to claim 1, characterized in that: The step of establishing an initial network structure according to the three-dimensional coordinates of the first frame point cloud in the laser point cloud data includes: According to the three-dimensional coordinates of the first frame point cloud, the maximum coordinates and the minimum coordinates of the first frame point cloud are calculated to obtain a spatial bounding box of the first frame point cloud; Calculate the average point spacing of the point cloud; Calculate the grid side length according to the average point spacing; The initial network structure is obtained by numbering the center coordinates according to the side length of the telescopic grid and the spatial bounding box.

3. The method for real-time de-stacking and fusion of multi-line structured light 3D reconstruction frame point cloud according to claim 1, characterized in that: The method of using subsequent frame point clouds in the laser point cloud data to classify point clouds and expand the network structure on the initial network structure to obtain a final network structure and determining a common grid according to the final network structure includes: Calculate the Euclidean minimum distance from the subsequent frame point cloud to the center point of the initial network structure; According to the Euclidean minimum distance of the center point, it is judged whether the subsequent frame point cloud exceeds the range of the initial network structure to obtain a judgment result. If the judgment result is yes, the current corresponding point cloud is stored. If the judgment result is no, the network structure of the current point cloud is expanded to obtain a public network.

4. The method for real-time de-stacking and fusion of multi-line structured light 3D reconstruction frame point cloud according to claim 1, characterized in that: The method of estimating the normal direction of the target point and eliminating the bidirectionality of the normal line to obtain an optimized normal vector according to the neighborhood point set and the principal component analysis method includes: Calculate the centroid of the preset point in the field point set according to the target point; Constructing a corresponding covariance matrix according to the centroid and the preset points in the domain point set; Decomposing the covariance matrix using singular values ​​to obtain eigenvalues ​​and eigenvectors; Obtaining a normal vector of the target point according to the eigenvalue and the eigenvector; The normal vector of the target point is bidirectionally eliminated to obtain an optimized normal vector.

5. The method for real-time de-stacking and fusion of multi-line structured light 3D reconstruction frame point cloud according to claim 1, characterized in that: The step of determining redundant points of the target point according to the optimized normal vector and removing the redundant points to obtain laser point cloud data after removal includes: The corresponding direction vector is calculated based on the optimized normal vector and the normal vector corresponding to the preset point in the domain point set; Calculate the angle between the direction vector and the optimized normal vector and the normal vector corresponding to the preset point to obtain a first angle and a second angle; Redundant points are determined according to the angle difference between the first angle and the second angle, and are eliminated to obtain laser point cloud data after elimination.

6. The method for real-time de-stacking and fusion of multi-line structured light 3D reconstruction frame point cloud according to claim 1, characterized in that: The step of establishing a cylindrical neighborhood of the target point according to the removed laser point cloud data comprises: Using the redundant point set corresponding to the domain point set to correct the normal vector of the target point in the laser point cloud data after the elimination, to obtain a corrected normal vector; Calculate the projection points of the target point along the approximate corrected normal vector in the positive and negative directions to determine the center coordinates of the upper and lower surfaces of the cylinder; The cylinder neighborhood is determined according to preset cylinder parameters and the center coordinates.

7. The method for real-time de-stacking and fusion of multi-line structured light 3D reconstruction frame point cloud according to claim 1, characterized in that: The step of iteratively optimizing the target point according to the cylindrical domain to obtain a fused point cloud includes: Identify redundant reference points; Optimizing and iterating the target point according to the redundant reference point to obtain a fusion point position; The fused point cloud is obtained according to the fused point positions.

8. The method for real-time de-stacking and fusion of multi-line structured light 3D reconstruction frame point cloud according to claim 7, characterized in that: The expression of the redundant reference point is: Among them, p i is the target point, p j is the neighborhood point, v ij is the direction vector, d i With d j They are v ij In p i and p j The forward projection length in the normal direction, p c Redundant reference point.

Images