A method for constructing a three-dimensional model of a building using laser scanning technology

Abstract

The application discloses a kind of building three-dimensional model construction method using laser scanning technology, it is related to building three-dimensional model construction technical field, comprising: obtaining the building point cloud of different scanning sites, adjacent building surface is formed by registration and building surface segmentation;According to the distance diagram of the recovery building point cloud of scanning site respectively, the building structure intersection line and actual observation line segment are formed according to building surface conversion position;Along the overall extension direction, space section is set, the scanning site of actual observation line segment is screened to cover each space section, and the building structure position is formed according to the space intersection of building structure intersection line and space section and fits final building structure intersection line;With final building structure intersection line as the common geometric boundary of adjacent building surface, building surface is refitted, and building three-dimensional model is constructed.The application realizes the distinction of actual observation part and mathematical extension part of building structure intersection line, unifies the connecting boundary of adjacent building surface, reduces the deviation, gap, overlap and intersection of connecting position.

Description

Technical Field

[0001] This invention relates to the field of architectural 3D model construction technology, and in particular to a method for constructing architectural 3D models using laser scanning technology. Background Technology

[0002] With the increasing demand for architectural spatial information in fields such as architectural surveying, digital preservation, and engineering renovation, 3D architectural models have been widely used for the digital recording of building geometry and spatial relationships. Laser scanning has the characteristics of non-contact measurement, wide coverage, and dense point cloud sampling, which can form architectural point clouds that reflect the spatial position of building surfaces. Wall corners, roof ridges, and the connection between walls and roofs are formed by two adjacent building surfaces, which are represented in 3D architectural models as the intersection lines between adjacent building surfaces and the grid boundaries formed along the intersection lines.

[0003] However, the existing methods still have two limitations: First, after merging the point clouds of buildings from multiple stations, the scanning coverage of each scanning station at the structural intersection is no longer retained, and the mathematical extension range of the intersection line of the local building plane space may exceed the actual observation range; Second, when adjacent building surfaces are fitted and grids are formed, the surface boundaries used may be offset due to differences in scanning direction and point cloud distribution, resulting in gaps, overlaps or intersections at the connection positions. Summary of the Invention

[0004] In view of the problems existing in the prior art, the present invention is proposed.

[0005] Therefore, this invention provides a method for constructing a 3D building model using laser scanning technology, which solves the problems of difficulty in distinguishing the actual observed part and the mathematically extended part of the intersection line of the building structure and the easy offset of the connection boundary between adjacent building surfaces when constructing a 3D building model using multi-station building point clouds.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a method for constructing a 3D building model using laser scanning technology. The method includes: acquiring building point clouds at different scanning stations; performing coordinate transformation, registration, and surface segmentation on the building point clouds to form adjacent building surfaces; reconstructing distance maps of the building point clouds according to each scanning station; extracting point clouds on both sides of the boundary based on the transformed positions of the building surfaces and fitting local building planes to form building structure intersection lines and actual observation line segments at each scanning station; forming an overall extension direction based on the building structure intersection lines; setting spatial sections along the overall extension direction; selecting scanning stations where the actual observation line segments cover each spatial section; obtaining the spatial intersection points of the building structure intersection lines and the spatial sections; forming building structure positions based on the spatial intersection points within the same spatial section; fitting the final building structure intersection lines based on the building structure positions formed along the overall extension direction; refitting the building surfaces using the final building structure intersection lines as the common geometric boundary of adjacent building surfaces; and constructing a 3D building model based on the refitted building surfaces.

[0007] As a preferred embodiment of the method for constructing a three-dimensional building model using laser scanning technology described in this invention, the step of acquiring building point clouds at different scanning stations includes: setting up control points and multiple scanning stations around the building, scanning the building facade, roof, wall corners, ridges, and the connection points between walls and roofs using a three-dimensional laser scanner at each scanning station to form building point clouds collected at different scanning stations, and recording the three-dimensional coordinates in the coordinate system of each scanning station, the original measurement distance, the scanning station identifier, and the original scanning arrangement information for each point cloud.

[0008] As a preferred embodiment of the method for constructing a 3D building model using laser scanning technology as described in this invention, the coordinate transformation, registration, and surface segmentation of the building point cloud include: forming a coordinate transformation relationship based on the spatial correspondence between each control point in the scanning station coordinate system and the unified building coordinate system; transforming the building point clouds collected from different scanning stations to the unified building coordinate system; and using an iterative nearest-point algorithm to perform fine registration of the building point clouds in the overlapping scanning area; denoising the finely registered building point clouds and calculating the local normal vectors of each point cloud point; segmenting the region based on the spatial adjacency relationship between point cloud points and the direction of the local normal vectors to form a building surface point cloud; and segmenting the building surface that can be represented using a plane. Plane fitting yields the building plan and its equations. The building surface point cloud is projected onto the building plan, and the building surface boundary is extracted along the edge of the projection point cloud's coverage area within the building plan. The median of the spatial distance between each point cloud point and its nearest point cloud point is taken as the inter-point distance of the building surface point cloud. The spatial intersection line is obtained based on the plane equations of the two non-parallel building plans. Boundary points from the two building surface boundaries that are no more than the inter-point distance of their respective building surface point clouds are extracted from the spatial intersection line. Boundary points belonging to the same continuous boundary section are projected onto the spatial intersection line according to their adjacent arrangement to form projection line segments. When the projection line segments formed by the two building surfaces overlap, the two building surfaces are considered as adjacent building surfaces.

[0009] As a preferred embodiment of the method for constructing a 3D building model using laser scanning technology as described in this invention, the step of restoring the distance map of the building point cloud according to the scanning station includes: processing adjacent building surfaces in groups; reading the building point cloud collected by each scanning station according to the scanning station identifier; arranging the point cloud points into a two-dimensional scanning grid according to the original scanning arrangement information; writing the original measured distance of each point cloud point into the grid position where the point cloud point is located, and simultaneously recording the 3D coordinates of the point cloud point in a unified building coordinate system and the building surface to which it belongs at the grid position; using the two-dimensional scanning grid that records the original measured distance as the distance map of the current scanning station; checking the positions of two directly adjacent grids that both record point cloud points along the scanning row and column directions of the distance map; when the two point cloud points belong to the two adjacent building surfaces being processed, recording the boundary between the two grid positions as the building surface transformation position; connecting the building surface transformation positions with common endpoints according to the scanning row and column relationship to form a continuous transformation section.

[0010] As a preferred embodiment of the method for constructing a 3D building model using laser scanning technology described in this invention, the step of extracting point clouds on both sides of the boundary based on the transformation position of the building surface and fitting local building planes to form the intersection lines of the building structures and the actual observed line segments at each scanning station includes: for each continuous transformation segment, reading point cloud points belonging to two adjacent building surfaces on both sides of the transformation position of the building surface, as well as point cloud points directly adjacent to the grid position of the point cloud points and belonging to the same building surface; summarizing and removing duplicate point cloud points according to their respective building surfaces to form point clouds on both sides of the boundary; when the point clouds on both sides of the boundary each include at least three non-collinear point cloud points, calculating the centroids of the point clouds on both sides of the boundary respectively, and forming a covariance matrix based on the 3D coordinate differences of each point cloud point relative to the centroid of the point cloud. Using the eigenvector corresponding to the smallest eigenvalue of the covariance matrix as the plane normal vector, local building planes are established, each passing through the centroid of the point cloud and with the normal direction being the plane normal vector. When two local building planes are not parallel, the spatial intersection line of the two local building planes is taken as the building structure intersection line of the current scanning station. The spatial midpoints of the two point cloud points on both sides of each building surface transformation position are obtained, and the spatial midpoints are orthogonally projected onto the building structure intersection line of the current scanning station to form projection points. According to the arrangement order of each projection point along the building structure intersection line, when at least two projection points with different positions are formed in the same continuous transformation segment, the part of the building structure intersection line between the first projection point and the last projection point is taken as the actual observation line segment, and the projection points in the actual observation line segment and the arrangement order of the projection points are retained.

[0011] As a preferred embodiment of the method for constructing a 3D building model using laser scanning technology described in this invention, the step of forming an overall extension direction based on the intersection lines of each building structure includes: processing adjacent building surfaces group by group, reading the intersection lines of building structures formed by different scanning stations and the actual observed line segments, forming a unit direction vector of the intersection line of the building structure based on the two endpoints of any actual observed line segment; multiplying the unit direction vector of the intersection line of the building structure at each scanning station with its own transpose vector to form a third-order direction matrix, calculating the arithmetic mean of the third-order direction matrices formed by different scanning stations according to the matrix elements to form an average matrix; performing eigenvalue decomposition on the average matrix, and normalizing the eigenvector corresponding to the largest eigenvalue to the unit direction vector of the overall extension direction.

[0012] As a preferred embodiment of the method for constructing a three-dimensional building model using laser scanning technology according to the present invention, the step of setting spatial sections along the overall extension direction, selecting scanning stations that cover each spatial section with actual observation line segments, and obtaining the spatial intersection points of the building structure intersection lines and spatial sections includes: obtaining the spatial distance between two adjacent projection points within the same actual observation line segment; taking the median value of each spatial distance at the same scanning station as the structural sampling interval; taking the median value of the structural sampling intervals at different scanning stations as the spatial section interval distance; selecting coordinate reference points from the endpoints of each actual observation line segment; taking the scalar projection of each actual observation line segment endpoint relative to the coordinate reference point along the overall extension direction as the endpoint position value; taking the range between the minimum and maximum values ​​of the position values ​​of each endpoint as the setting range of the spatial sections; setting spatial sections sequentially according to the spatial section interval distance starting from the minimum position value; the normal direction of each spatial section is consistent with the overall extension direction; calculating the spatial intersection points of the building structure intersection lines and each spatial section at each scanning station; when the spatial intersection point is located between the two endpoints of any actual observation line segment at the current scanning station, the current scanning station is taken as the scanning station where the actual observation line segment covers the current spatial section and the spatial intersection point is retained.

[0013] As a preferred embodiment of the method for constructing a 3D building model using laser scanning technology as described in this invention, the step of forming building structure positions based on spatial intersections within the same spatial section and fitting the final building structure intersection line based on the building structure positions formed along the overall extension direction includes: when the same spatial section retains spatial intersections from no less than three different scanning stations, and the spatial intersections are not all located on the same straight line, the spatial position with the smallest sum of Euclidean distances among the spatial intersections is taken as the building structure position of the current spatial section; the building structure positions are arranged according to the overall extension direction, and when at least two different building structure positions are formed, the building structure positions are calculated. The arithmetic mean of the three-dimensional coordinates of the structural positions is used to form the average position of the building structure. Difference vectors between the three-dimensional coordinates of each structural position and the average position are then generated. Each difference vector is multiplied by its transpose, and the arithmetic mean of all product matrices is calculated element-wise to form the position covariance matrix of the building structure. The position covariance matrix is ​​then subjected to eigenvalue decomposition. The eigenvector corresponding to the largest eigenvalue is normalized to the unit direction vector of the final structural intersection line. The spatial straight line passing through the average position of the structural positions and extending along the unit direction vector of the final structural intersection line is taken as the final structural intersection line.

[0014] As a preferred embodiment of the method for constructing a 3D building model using laser scanning technology described in this invention, the refitting of the building surface using the final building structure intersection line as the common geometric boundary of adjacent building surfaces includes: orthogonally projecting the first and last building structure positions arranged along the overall extension direction to the final building structure intersection line, and using the portion of the final building structure intersection line between the two projection points as the intersection line constraint range; reading the building surface point clouds of two adjacent building surfaces respectively, and taking the spatial midpoint of the intersection line constraint range as the plane fitting reference point; for each building surface, establishing a candidate plane that passes through the plane fitting reference point and whose normal direction is perpendicular to the unit direction vector of the final building structure intersection line, selecting the candidate plane with the smallest sum of squared orthogonal distances from each point cloud in the building surface point cloud to the candidate plane, and forming the refitted building surface containing the final building structure intersection line respectively; and using the intersection line constraint range as the common geometric boundary of the two refitted building surfaces.

[0015] As a preferred embodiment of the method for constructing a 3D building model using laser scanning technology as described in this invention, the step of constructing a 3D building model based on the refitted building surface includes: orthogonally projecting the point clouds of the building surfaces of the two building surfaces onto the refitted building surface; orthogonally projecting each boundary point in the boundary of the building surface extracted when forming an adjacency relationship onto the refitted building surface to form projected boundary points, retaining the correspondence between each boundary point and the projected boundary points, and connecting the projected boundary points according to the original adjacent arrangement of each boundary point to form the refitted building surface boundary; reading each spatial section within the intersection constraint range, obtaining the spatial intersection points of the final building structure intersection line and each spatial section respectively, arranging the projection points of the first and last building structure positions on the final building structure intersection line with each spatial intersection point according to the overall extension direction, removing points with overlapping spatial positions, and forming common boundary vertices of the two building surfaces; obtaining the projection boundary points corresponding to the continuous boundary portion used when the two building surfaces form an adjacency relationship according to the correspondence between each boundary point and the projected boundary points; orthogonally projecting the obtained projection boundary points onto the final building structure intersection line, removing the projection points of the intersection line located at the intersection point; Within the line constraint range, projected boundary points are selected, and common boundary vertices are written into the boundary portion where the removed projected boundary points are located, forming the updated building surface boundaries of the two building surfaces. The orthogonal projection positions of the building surface point cloud on the refitted building surface are used as the internal vertices of the building surface, and the updated building surface boundaries are used as the triangulation range. The lines connecting adjacent common boundary vertices are used as constraint edges, and constrained Delaunay triangulation is performed on the two building surfaces to form triangular meshes of the two building surfaces. The triangular meshes of the two building surfaces use common boundary vertices with the same 3D coordinates and vertex numbers, and jointly reference the constraint edges between adjacent common boundary vertices. Triangular meshes are constructed on adjacent building surfaces that have not formed the final building structure intersection line using the already formed building plane and building surface boundaries. Based on the retained building surface point cloud, curved triangular meshes are reconstructed for building surfaces that cannot be represented by a plane. The triangular meshes and curved triangular meshes of each building surface are summarized, and the mesh vertices that coincide in spatial position and are located at the connection position of the same building surface are merged. The vertex arrangement order of the triangular facets is unified according to the original normal direction of each building surface to form the 3D building model.

[0016] The beneficial effects of this invention are as follows: by constraining the scanning stations with actual observation line segments to participate in the formation of the spatial cross-section building structure position, the building structure position is formed according to the spatial intersection points within the same spatial cross-section and the final building structure intersection line is fitted, thus realizing the distinction between the actual observation part and the mathematical extension part of the building structure intersection line, reducing the influence of the extended part of the intersection line on the building structure position; by using the final building structure intersection line as the common geometric boundary of adjacent building surfaces, and using common boundary vertices and constraint edges in the triangular mesh, the connection position of adjacent building surfaces is unified, reducing the boundary offset, gaps, overlaps and intersections of the connection positions of the three-dimensional building model. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a method for constructing 3D architectural models using laser scanning technology.

[0019] Figure 2 A flowchart for generating the intersection lines of building structures and the actual observation lines at each scanning station.

[0020] Figure 3 A flowchart for fitting the final intersection line of the building structure.

[0021] Figure 4 A flowchart for refitting building surfaces and constructing a 3D model of the building.

[0022] Figure 5 This is a map showing the actual distribution of the observed line segments and building structures.

[0023] Figure 6 This is a diagram showing the connection of the common geometric boundary of adjacent building surfaces. Detailed Implementation

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0025] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0026] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0027] Reference Figures 1-6 As one embodiment of the present invention, this embodiment provides a method for constructing a three-dimensional building model using laser scanning technology, comprising the following steps: Architectural 3D models need to accurately reflect the spatial location and connection relationships of building structures such as walls, roofs, corners, and ridges. Laser scanning can directly obtain the 3D coordinates and measurement distances of building surfaces in a non-contact manner. Compared with relying on manual measurement or image features to infer the building's spatial structure, it is more suitable for quickly obtaining continuous 3D geometric information of complex buildings. Since wall bends, eaves, and roofs with different orientations can limit the visible range of a single scanning position, actual modeling usually requires multi-station laser scanning from different locations, and the building point clouds collected from each scanning station are converted to a unified building coordinate system.

[0028] While multi-station scanning can supplement obscured or poorly observed building surfaces, the different scanning distances, observation directions, and occlusion relationships between different scanning stations and the same building structure result in variations in point cloud density and spatial distribution at the same corner, ridge, or wall-to-roof connection point. If the point clouds from each scanning station are directly mixed for building surface fitting after registration, scanning stations with more point clouds will participate in the fitting with more measurement points. Positional deviations formed by a single scanning direction at the building edge may also be carried over into the fitting results, leading to situations where each building surface fits well individually, but the actual connection position is offset. Based on this, this embodiment forms adjacent building surfaces according to the spatial connection relationship between building surfaces, retains the original observation relationship corresponding to each scanning station, and allows each scanning station to form independent geometric results for the same building structure location based on the actual collected point clouds. The intersection lines of the building structures are restored using the independent observation results of different scanning stations for the same building structure location, and the restored intersection lines constrain the 3D construction of adjacent building surfaces, thereby reducing the positional deviations caused by a single scanning direction from being directly incorporated into the 3D building model.

[0029] S1. Obtain building point clouds at different scanning stations, perform coordinate transformation, registration, and building surface segmentation on the building point clouds to form adjacent building surfaces.

[0030] Furthermore, control points are set up around the building to be modeled in 3D. These control points are spatial reference points with known spatial coordinates that can be identified by a 3D laser scanner. Multiple scanning stations are set up according to the overall size of the building, the shape of the facade, and the site occlusion. A 3D laser scanner is set up at each scanning station to scan the building facade and roof, and to cover the corners of the walls, the ridges, and the connection between the walls and the roof, forming a building point cloud collected by different scanning stations.

[0031] Furthermore, for the building point cloud collected by each scanning station, each point cloud point records at least its three-dimensional coordinates in the scanning station coordinate system, the original measurement distance, and the scanning station identifier. At the same time, it retains the original scanning arrangement information of the point cloud point in the current scanning station. The original measurement distance is the measurement distance from the distance measuring center of the three-dimensional laser scanner to the corresponding building surface position of the current point cloud point.

[0032] Among them, the original scan arrangement information reflects the adjacent arrangement relationship between point cloud points when the same scanning station continuously scans the surface of a building.

[0033] Furthermore, when the 3D laser scanner outputs the scan row number and scan column number, it saves the scan row number and scan column number corresponding to each point cloud point; when the 3D laser scanner records the scan position according to the horizontal scan angle and vertical scan angle, it saves the horizontal scan angle and vertical scan angle corresponding to each point cloud point, and obtains the horizontal angular resolution and vertical angular resolution used by the current scanning station, so that each point cloud point can restore the original arrangement relationship according to the actual scanning order.

[0034] Furthermore, the spatial coordinates of each control point in the unified building coordinate system are obtained, and the control points are identified in the building point cloud collected by each scanning station. For the current scanning station, at least three non-collinear control points that can obtain spatial coordinates in the scanning station coordinate system and the unified building coordinate system are selected. Based on the spatial position correspondence of each control point in the two coordinate systems, the rotation and translation relationship of the scanning station coordinate system relative to the unified building coordinate system is obtained to form a coordinate transformation relationship. The building point cloud collected by each scanning station is transformed to the unified building coordinate system according to the coordinate transformation relationship.

[0035] The unified building coordinate system refers to a spatial coordinate system established based on the engineering coordinate benchmark used in building surveying, which is used to uniformly represent the location of building point clouds collected by different scanning stations.

[0036] Furthermore, for two scanning stations with overlapping scanning areas on the building surface, building point clouds within the overlapping scanning areas are extracted respectively, and fine registration is performed using the Iterative Closest Point (ICP) algorithm. During the iteration process, a point cloud matching relationship is established based on the spatial distance between the two sets of overlapping point clouds, and the rotation and translation amount that reduces the spatial distance between the matching points is calculated. The building point cloud position is updated based on the rotation and translation amount until the overlapping point cloud registration is completed. The spatial position of the building point cloud of the corresponding scanning station in the unified building coordinate system is corrected based on the fine registration result.

[0037] It should be noted that control point coordinate transformation brings building point clouds collected by different scanning stations into the same spatial coordinate reference. However, scanner setup, control point measurement, and discrete sampling of point clouds may still cause local positional differences in overlapping point clouds on the same building surface. Therefore, further fine registration is performed using the overlapping point clouds themselves. Coordinate transformation and fine registration change the spatial position of point cloud points in a unified building coordinate system, without changing the scanning station identification and original scanning arrangement information. The scanning station identification distinguishes the actual scanning source of the point cloud points, and the original scanning arrangement information retains the adjacency relationship of each point cloud point in the same scanning station during the actual scanning process. Therefore, after spatial registration is completed, it is still possible to read the independent scanning data formed by different scanning stations.

[0038] Furthermore, the registered building point cloud is denoised to remove isolated points that are spatially discrete from the surrounding point cloud. For each retained point cloud point, the nearest point cloud point is found according to the three-dimensional spatial distance. A covariance matrix is ​​formed based on the three-dimensional coordinates of the found point cloud points, and the eigenvector corresponding to the smallest eigenvalue of the covariance matrix is ​​used as the local normal vector of the current point cloud point.

[0039] Among them, the local normal vector reflects the spatial orientation of the building surface near the point cloud point. The positive and negative directions of the local normal vector indicate the same building surface orientation. The building surface orientation of adjacent points inside the same wall or roof surface is usually similar, while the building surface orientations on both sides of the wall corner, roof ridge, and the connection between the wall and the roof are different. Therefore, it is possible to distinguish different building surfaces by combining the spatial adjacency relationship between point cloud points.

[0040] Furthermore, the building point cloud is segmented into regions based on the spatial adjacency relationship between point cloud points and the local normal vector direction. When comparing the local normal vector directions of spatially adjacent points, the smaller value between the angle between two local normal vectors and the angle formed by reversing one of the local normal vectors is taken to form the building surface point cloud corresponding to different building surfaces. Then, the building surfaces that can be represented by a plane are fitted with a plane to obtain the building plane and plane equation corresponding to each building surface.

[0041] Furthermore, for each building surface that can be represented by a plane, the point cloud of the building surface is projected onto the corresponding building plane, and the boundary of the building surface is extracted along the edge of the coverage area of ​​the projected point cloud in the building plane. At the same time, the spatial distance between each point cloud point and the nearest point cloud point is calculated, and the median value of each spatial distance is taken as the point distance between the current building surface point cloud. The point distance reflects the actual sampling density of the current building surface point cloud. Using the median value can reduce the impact of local point cloud missing or abnormally dense points on the overall sampling spacing judgment.

[0042] Furthermore, for any two non-parallel building planes, spatial lines simultaneously located within the two building planes are obtained based on their plane equations. The extension direction of these spatial lines is perpendicular to the normal vectors of both building planes. These obtained spatial lines are taken as the spatial intersection line of the two building planes. Boundary points whose distance to the spatial intersection line is no greater than the distance between the point clouds of their respective building surfaces are extracted from the boundaries of the two building surfaces. According to the adjacent arrangement of the boundary points in the building surface boundaries, the boundary points belonging to the same continuous boundary portion are projected onto the spatial intersection line. The spatial intersection line portion between the first and last projection points of the same continuous boundary portion is taken as the projection line segment. When both building surfaces can form projection line segments, and the projection line segments of the two building surfaces have overlapping line segments, the two building surfaces are considered as adjacent building surfaces.

[0043] Adjacent building surfaces include two walls connected to form a wall corner, two sloping roofs connected to form a roof ridge, and walls connected to a flat roof or sloping roof to form a straight boundary. If the building planes corresponding to two building surfaces can intersect in space, but the boundaries of the building surfaces do not extend to the spatial intersection line, or the two building surfaces do not form a common covering line segment along the spatial intersection line direction, then the two building surfaces are not considered as adjacent building surfaces.

[0044] It should be noted that the existence of a spatial intersection line between two building plans only indicates that the two building plans can intersect after being extended infinitely, and does not indicate that the two building surfaces have a connection relationship in the actual building. Therefore, it is also necessary to check the distance between the boundaries of the two building surfaces and the spatial intersection line, and to determine whether the projected line segments formed by the two building surfaces along the spatial intersection line overlap. At the same time, the laser scan forms a discrete point cloud, and the boundary points of the building surfaces usually do not fall exactly on the calculated spatial intersection line. Therefore, it is necessary to determine whether the boundary of the building surface has extended to the spatial intersection line based on the distance between the points in the current building surface point cloud, so that the boundary judgment is adapted to the actual sampling density of the current building surface point cloud.

[0045] It should also be noted that for curved roofs, architectural decorative curved surfaces, and building surfaces where connection points cannot be represented by a local planar representation, the corresponding building surface point cloud is retained and processed according to the conventional curved surface reconstruction method, and the corresponding connection points are not represented by the intersection lines of the building planar space.

[0046] Furthermore, architectural photographic images of the 3D model to be constructed are acquired, along with the camera's in-camera parameters and its extrinsic parameters relative to a unified architectural coordinate system. Based on the extrinsic parameters, spatial points in the unified architectural coordinate system are transformed to the camera coordinate system. Then, based on the camera's in-camera parameters and perspective projection relationships, the pixel positions of the spatial points in the architectural photographic images are calculated, establishing a spatial calibration relationship between the architectural photographic images and the unified architectural coordinate system.

[0047] S2. Restore the distance map of the building point cloud according to the scanning station, extract the point cloud on both sides of the boundary according to the conversion position of the building surface and fit the local building plane to form the intersection line of the building structure and the actual observation line segment of each scanning station.

[0048] Furthermore, adjacent building surfaces are processed group by group, and the building point clouds collected by each scanning station are read according to the scanning station identifier. When the current scanning station collects point cloud points from two adjacent building surfaces at the same time, the building point cloud collected by the current scanning station is read. When the current scanning station does not collect point cloud points from two adjacent building surfaces at the same time, the current scanning station is not processed.

[0049] Furthermore, for the current scanning site, when the point cloud points store scan row numbers and scan column numbers, a two-dimensional scanning grid is established according to the scan row numbers and scan column numbers, and each point cloud point is arranged to the grid position where its own scan row number and scan column number are located; when the point cloud points store horizontal scanning angles and vertical scanning angles, scan column positions and scan row positions are formed according to the horizontal scanning angles and vertical scanning angles in ascending order, combined with the horizontal angular resolution and vertical angular resolution respectively, and each point cloud point is arranged to the formed scan row positions and scan column positions, thereby organizing the building point cloud according to the arrangement relationship when the current scanning site actually scans the building surface.

[0050] Furthermore, in the two-dimensional scanning grid, the original measured distance of each point cloud point is written into the grid position where the point cloud point is located, while retaining the three-dimensional coordinates of the point cloud point in the unified building coordinate system and the building surface to which it belongs; the grid positions where no point cloud points are recorded remain empty, and the original measured distances recorded by each grid position are arranged according to the scanning row and column relationship to form the distance map of the current scanning station.

[0051] Among them, the distance map reflects the distance measurement results of the building surface formed by the current scanning station according to the actual scanning row and column relationship. The grid positions of adjacent points in the same row or column that are all recorded point cloud points represent the adjacent measurement positions during the continuous scanning process of the 3D laser scanner. At the same time, the 3D coordinates of the point cloud points in the unified building coordinate system and the building surface to which they belong can be read through the grid positions.

[0052] Furthermore, along the scan row and scan column directions of the distance map, the positions of two directly adjacent grids that both record point cloud points are checked sequentially, and the building surfaces to which the two point cloud points belong are read. When the two point cloud points belong to the two adjacent building surfaces being processed, the boundary between the two grid positions is recorded as the building surface transformation position, and the point cloud points on both sides of the transformation position and the scan row and column positions are recorded. When the two point cloud points belong to the same building surface, or when the point cloud points in any grid position do not belong to the two adjacent building surfaces being processed, no building surface transformation position is formed.

[0053] Among them, the building surface transition position represents the scanning boundary where the scanning position moves from one adjacent building surface to another when the current scanning station continuously collects building surfaces according to the actual scanning row and column relationship. The building surface transition position itself is not used as a three-dimensional spatial point, but is represented by the point cloud points actually collected on both sides of the transition position.

[0054] It should also be noted that the grid adjacency relationship in the distance map reflects the acquisition order between adjacent laser scanning positions in the same scanning station. When buildings obstruct or front and back surfaces alternate, two building surfaces that are not connected in real space may also form adjacent grid positions in the distance map. Therefore, the building surface conversion position is only searched between two building surfaces that have already formed an adjacency relationship, and the actual connection relationship between building surfaces is not re-determined using the grid adjacency relationship in the distance map. At the same time, the building surfaces on both sides of the wall corner, roof ridge, and the connection position between the wall and the roof have different spatial orientations, but the original measurement distance of the continuous scanning position may change continuously or only change slightly. Therefore, the magnitude of the change in the original measurement distance between two adjacent grid positions is not used as the condition for the formation of the building surface conversion position. Instead, it is determined based on whether the building surface to which the grid position of the two recorded point cloud points belongs has changed.

[0055] Furthermore, the transformation positions of building surfaces are checked according to the scanning row and column relationship in the distance map. When two transformation positions of building surfaces have a common endpoint on the boundary of the two-dimensional scanning grid, the two transformation positions of building surfaces are connected. And the building surface transformation positions that can be connected to each other are connected in the same way to form a continuous transformation section. When there are empty grid positions between building surface transformation positions or grid positions belonging to other building surfaces, the connection is not made across the intermediate positions. When the current scanning station has not formed a building surface transformation position, the building structure intersection line and the actual observation line segment are not formed using the current scanning station.

[0056] Furthermore, for each continuous transformation segment, the point cloud points belonging to two adjacent building surfaces on both sides of the transformation position of the building surface are read, and the point cloud points directly adjacent to the grid positions of these point cloud points and belonging to the same building surface are read. The three-dimensional coordinates of the read point cloud points in a unified building coordinate system are obtained. The point cloud points belonging to the same building surface in each continuous transformation segment of the current scanning station are summarized and duplicate point cloud points are removed to form point clouds on both sides of the boundary.

[0057] Furthermore, the point clouds on both sides of the boundary are processed separately, and the average position of the three-dimensional coordinates of the point cloud points is calculated as the centroid of the point cloud. A covariance matrix is ​​formed based on the difference in three-dimensional coordinates of each point cloud point relative to the centroid of the point cloud. The covariance matrix is ​​eigenvalued, and the eigenvector corresponding to the smallest eigenvalue is taken as the plane normal vector. A local building plane passing through the centroid of the point cloud and whose normal direction is the plane normal vector is established, and the equation of the local building plane is formed. The local building planes of two adjacent building surfaces under the current scanning station are obtained.

[0058] Furthermore, when there are fewer than three point cloud points on either side of the boundary, or when all point cloud points are located on the same straight line, a local building plane is not formed; when local building planes are formed on both sides of the boundary and the two local building planes are not parallel, a spatial straight line that is simultaneously located in the two local building planes is obtained according to the plane equations of the two local building planes, and the spatial straight line is used as the building structure intersection line of the current scanning station; when the two local building planes are parallel, the building structure intersection line is not formed using the current scanning station.

[0059] One or two points in space cannot uniquely define a plane, and three points on the same straight line also cannot uniquely define a plane. Therefore, a local building plan with a unique normal direction can only be formed when there are at least three non-collinear point cloud points on one side of the boundary.

[0060] It should be noted that the building surface conversion position is located at the scanning boundary of two building surfaces. When the laser spot is close to the corner of the wall, the ridge, or the connection between the wall and the roof, it may cover the building surfaces on both sides of the boundary at the same time. Reflection at the building edge may also cause the measurement position of a small number of point cloud points to deviate from the actual building structure position. Therefore, the building structure intersection line is not directly formed by using the individual point cloud points on both sides of the building surface conversion position. Instead, the local building plane is fitted according to the actual building surface point cloud collected on both sides of the boundary, and the spatial intersection relationship of the two local building planes is used to form the building structure intersection line. This ensures that the positional deviation of a small number of building edge measurement points will not directly determine the spatial position and extension direction of the building structure intersection line.

[0061] It should also be noted that the connection relationship between building surfaces is determined based on the point cloud and building plan after coordinate transformation and fine registration. The local building plan is formed only by fitting the point cloud actually collected by the current scanning station on both sides of the building boundary. The two use different point cloud ranges. Using the point cloud of a single scanning station to form the local building plan can preserve the independent geometric observation of the same building structure location by different scanning stations, and avoid mixing the point clouds of different scanning stations again, which would weaken the observation differences between the scanning stations.

[0062] Furthermore, the transformation positions of the building surface in the continuous transformation section are processed one by one. Two point cloud points on both sides of the transformation position are read. The spatial midpoint is obtained based on the three-dimensional coordinates of the two point cloud points in the unified building coordinate system. The spatial midpoint is orthogonally projected onto the intersection line of the building structure of the current scanning station to form a projection point.

[0063] The building surface transformation position is located between two adjacent scanning positions. The point cloud points on both sides of the transformation position come from the building surfaces on both sides of the boundary. Therefore, the spatial midpoint of the two point cloud points can represent the approximate spatial location where the building surface transformation occurs at the current scanning position. Projecting the spatial midpoint onto the building structure intersection line is only used to reflect the actual range of the building structure intersection observed by the current scanning station, without changing the position and extension direction of the building structure intersection line formed by the point clouds on both sides of the boundary.

[0064] Furthermore, the projection points are arranged according to their order along the intersection lines of the building structure. When at least two projection points with different positions are formed in the same continuous transition section, the intersection line between the first and last projection points is taken as the actual observation line segment. When there are multiple discontinuous continuous transition sections at the current scanning station, actual observation line segments are formed separately, without connecting the intersection lines of the building structure that lack the support of the building surface transition position between different actual observation line segments. When a continuous transition section forms only one projection point, no actual observation line segment is formed. The projection points and their arrangement order in each actual observation line segment are retained.

[0065] It should be noted that the intersection line of the building structure formed by two local building planes can mathematically extend continuously in both directions. However, the current scanning station only collected point clouds on both sides of the intersection within the actual area where the building surface transformation occurs. Therefore, the intersection line of the building structure that exceeds the actual observation line segment is only a mathematical extension of the local building plane and cannot indicate that the current scanning station has observed the corresponding building structure position. Retaining the actual observation line segment can avoid taking the extended part of the intersection line of the building structure in the building occlusion, scanning blind zone, or uncollected area as the actual observation result of the current scanning station.

[0066] S3. Based on the intersection lines of each building structure, form an overall extension direction, set up spatial sections along the overall extension direction, select scanning stations that cover each spatial section with actual observation line segments, obtain the spatial intersection points of the building structure intersection lines and spatial sections, form the building structure positions based on the spatial intersection points within the same spatial section, and fit the final building structure intersection lines based on the building structure positions formed along the overall extension direction.

[0067] Furthermore, adjacent building surfaces are processed group by group, and the structural intersection lines and actual observation line segments formed by different scanning stations are read. When the current scanning station has formed both the structural intersection line and the actual observation line segment, two endpoints are read from any actual observation line segment. Since the actual observation line segment is located on the structural intersection line, one endpoint is taken as the first spatial point and the other endpoint as the second spatial point, with the first spatial point as the base point of the structural intersection line. A difference vector between the three-dimensional coordinates of the second spatial point and the first spatial point is formed, and then normalized according to the Euclidean length of the difference vector to obtain the unit direction vector of the structural intersection line, expressed as: ; In the formula, For the first Unit direction vector of the intersection line of the building structure at each scanning site; To scan the site sequence number; For the first The three-dimensional coordinate vector of the first spatial point on the intersection line of the building structure of each scanning site; For the first The three-dimensional coordinate vector of the second spatial point on the intersection line of the building structure of each scanning site; This represents the Euclidean length of the vector.

[0068] The first and second spatial points are used only to establish the directional representation and straight line parameter representation of the intersection line of the building structure, and do not represent the fixed positive and negative directions of the intersection line of the building structure; either endpoint can be used as the first spatial point and the other endpoint as the second spatial point. Interchanging the two endpoints only changes the positive and negative directions of the unit direction vector and the positive and negative directions of the directed distance when representing the spatial position along the intersection line of the building structure, and does not change the spatial position of the intersection line of the building structure itself.

[0069] Furthermore, the unit direction vector of the intersection line of the building structures at each scanning site is multiplied by its own transpose to form a third-order direction matrix, represented as: ; In the formula, For the first The third-order orientation matrix of each scanning station; Transpose a vector or matrix; Furthermore, the third-order directional matrices formed by different scanning stations are arithmetically averaged element-wise to form an average matrix, which is represented as: ; In the formula, This is the average matrix of the direction matrices of each scanning station; This represents the number of scanning stations that form the intersection line of the building structure and the actual observation line segment on the surfaces of two adjacent buildings.

[0070] Furthermore, eigenvalue decomposition is performed on the average matrix to obtain the eigenvector corresponding to the largest eigenvalue. This eigenvector is then normalized to a unit vector. The spatial direction of the unit vector is used as the overall extension direction of the two adjacent building surfaces, expressed as: ; In the formula, The largest eigenvalue of the average matrix; The unit direction vector of the overall extension direction is obtained by normalizing the eigenvector corresponding to the largest eigenvalue of the average matrix.

[0071] It should be noted that the same spatial line can be represented by two unit direction vectors with opposite directions. When directly averaging the unit direction vectors of different scanning stations, the unit direction vectors with opposite directions may cancel each other out. After multiplying the unit direction vector with its own transpose, the positive unit direction vector and the negative unit direction vector form the same third-order direction matrix. Therefore, it is not necessary to unify the positive and negative directions of the intersection lines of each building structure in advance. The common overall extension direction can be obtained based on the intersection lines of the building structures formed independently by different scanning stations. The selection order of the first and second spatial points will not change the third-order direction matrix and the final overall extension direction. The overall extension direction is used to uniformly represent the common extension axis of the intersection lines of each building structure. The positive or negative direction only affects the positive and negative position value and the arrangement direction when representing the spatial position along the extension axis. The subsequent spatial section setting, spatial intersection point screening and building structure position arrangement are all carried out in accordance with the currently formed overall extension direction.

[0072] Furthermore, the projection points used to form each actual observation line segment are read, and the actual observation line segments are processed one by one according to the arrangement order of the projection points along the intersection line of the building structure; the spatial distance between the three-dimensional coordinates of two adjacent projection points within the same actual observation line segment is obtained, but the spatial distance between projection points of different actual observation line segments is not calculated; the spatial distances obtained from the same scanning station are summarized, and the median value of each spatial distance is taken as the structural sampling interval of the current scanning station; then the median value of the structural sampling intervals of different scanning stations is taken as the spatial cross-sectional interval distance.

[0073] The structural sampling interval reflects the overall sampling density between projection points when the building structure is actually observed at the current scanning station. Due to differences in scanning distance, scanning direction, and incident angle on the building surface between different scanning stations and the building structure, the spatial distribution density of projection points formed by the same building structure may differ at different scanning stations. At the same time, the edges of the building structure, occluded areas, or locally densely scanned areas may produce spatial distances that are too large or too small. Therefore, the median value of each spatial distance is used to form the structural sampling interval of the current scanning station, which can reduce the impact of a few abnormal spatial distances on the overall sampling density judgment. Then, the spatial cross-section interval distance is formed based on the median value of the structural sampling interval of different scanning stations, so that multiple scanning stations can participate in the spatial cross-section setting according to their own actual sampling characteristics, avoiding the influence of a single scanning station on the spatial cross-section setting due to a large number of projection points or locally dense sampling.

[0074] Furthermore, one endpoint is selected from the endpoints of each actual observed line segment as a coordinate reference point. For each endpoint of the actual observed line segment, a difference vector is formed between the three-dimensional coordinates of the endpoint and the three-dimensional coordinates of the coordinate reference point. The difference vector is then scalar-projected along the overall extension direction, and the resulting projection length represents the position of the endpoint along the overall extension direction, expressed as: ; In the formula, For the first The position values ​​of the endpoints of each actual observed line segment along the overall extension direction; The actual observation line segment endpoint number; For the first The three-dimensional coordinate vectors of the endpoints of each actual observed line segment; The three-dimensional coordinate vector is the coordinate reference point.

[0075] Furthermore, the minimum and maximum values ​​of the endpoint positions of each actual observed line segment are taken, and the range between the minimum and maximum position values ​​is used as the setting range of the spatial cross-section. Starting from the minimum position value, spatial cross-section positions are formed sequentially along the overall extension direction according to the spatial cross-section interval distance. The process stops when the spatial cross-section interval distance exceeds the maximum position value. When the last formed spatial cross-section position does not coincide with the maximum position value, the maximum position value is added as the last spatial cross-section position. Spatial cross-sections are set according to the order of their positions along the overall extension direction, and the normal direction of each spatial cross-section is consistent with the overall extension direction, as shown below: ; In the formula, is the three-dimensional coordinate vector of any point in space within the spatial section; For spatial section number; For the first The positional value of each spatial section along the overall extension direction.

[0076] Furthermore, taking the first spatial point as the base point of the structural intersection line, the directed distance calculated from the first spatial point along the unit direction vector of the structural intersection line is used as the position parameter of the structural intersection line. This allows each spatial point on the structural intersection line to be represented based on the base point, the unit direction vector, and the directed distance. The structural intersection line is thus represented as: ; In the formula, For the first The location parameters are located on the intersection of the building structure at each scanning site. Represents the three-dimensional coordinate vector of a point in space; It is the directed distance calculated from the first spatial point along the unit direction vector of the intersection line of the building structure, and its dimensions are consistent with those of the three-dimensional coordinates.

[0077] Furthermore, each spatial section is processed individually. The spatial points on the intersection line of the building structure are substituted into the plane equation of the current spatial section to establish the equations that satisfy the plane equation of the current spatial section. The directed distance between the intersection line of the building structure and the current spatial section and the first spatial point is then calculated, expressed as: ; In the formula, For the first The intersection of the building structure at the first scanning site with the first The directed distance calculated from the first spatial point along the unit direction vector of the intersection line of the building structure when two spatial sections intersect.

[0078] Furthermore, by substituting the directed distance when the building structure intersection line intersects with the current spatial section into the parametric equation of the building structure intersection line, we obtain the spatial position that is simultaneously located within both the building structure intersection line and the current spatial section, forming the spatial intersection point of the building structure intersection line and the spatial section, represented as: ; In the formula, For the first The intersection of the building structure at the first scanning site with the first The three-dimensional coordinate vector of the spatial intersection point of a spatial section.

[0079] When the overall extension direction is perpendicular to the unit direction vector of the intersection line of the building structure at the current scanning station, the intersection line of the building structure is parallel to or located within the spatial section, and cannot form a unique spatial intersection point. Therefore, the current scanning station is not used to participate in the formation of the building structure position of the current spatial section.

[0080] Furthermore, for scanning stations that can form a unique spatial intersection, it is checked whether the spatial intersection is located between the two endpoints of any actual observation line segment of the current scanning station; when the spatial intersection is located within any actual observation line segment, the current scanning station is used as the scanning station covering the current spatial section as the actual observation line segment and the spatial intersection is retained; when the spatial intersection is located outside all actual observation line segments, the extended portion of the building structure intersection line located outside the actual observation line segment is not used to participate in the formation of the building structure position of the current spatial section.

[0081] It should be noted that the intersection line of the building structure is formed based on the spatial intersection relationship of two local building planes. Mathematically, it can extend beyond the actual scanning range. However, the actual observation line segment limits the range of the building structure supported by real scanning data at the current scanning station. Therefore, the current scanning station only participates in the formation of the building structure position of the current spatial section when the spatial intersection point of the building structure intersection line and the spatial section is located within the actual observation line segment. The same scanning station retains at most one spatial intersection point within the same spatial section, so that different scanning stations participate in the formation of the building structure position according to independent scanning sources, avoiding repeated participation by a single scanning station due to the large number of point cloud points.

[0082] Furthermore, each spatial section is processed individually, and the spatial intersections formed by different scanning stations covering the current spatial section are summarized. The retained spatial intersections are then arranged according to the scanning station number. When the same spatial section retains spatial intersections from no fewer than three different scanning stations, and not all spatial intersections are located on the same straight line, the spatial position in three-dimensional space that minimizes the sum of the Euclidean distances of all spatial intersections within the current spatial section is determined. This determined spatial position is taken as the architectural structural position of the current spatial section, expressed as: ; In the formula, For the first Three-dimensional coordinate vectors of the architectural structural position of each spatial section; To obtain the three-dimensional coordinate vector of any candidate spatial position during the process of determining the location of the building structure; The sequence number of the scan sites, retained after being arranged according to their serial numbers; For the first The number of spatial intersections of different scanning stations preserved in each spatial section; The first one, arranged according to the scan site number The intersection of the building structure at the first scanning site with the first Three-dimensional coordinate vectors of the spatial intersection points of a spatial section; This represents the candidate spatial location that minimizes the sum of the Euclidean distances between all spatial intersections.

[0083] The geometric median is the spatial location that minimizes the sum of Euclidean distances between all spatial intersections within the same spatial cross-section, and can be obtained using the Weiszfeld iterative algorithm. The building structure location is formed using spatial intersections from at least three different scanning stations because when only two spatial intersections are used, the sum of distances from any point between the two intersections is the same, making it impossible to uniquely determine the building structure location based on the minimum distance principle. When multiple spatial intersections are formed using at least three different scanning stations, these intersections can collectively constrain the building structure location. Even when multiple spatial intersections are all located on the same straight line, spatial location degradation may still occur, therefore the building structure location for the current spatial cross-section is not formed.

[0084] It should also be noted that the intersection lines of building structures formed by different scanning stations may have slight spatial misalignment due to scanning distance, scanning direction, and measurement deviations of building edges. By obtaining the spatial intersection points of each building structure intersection line in the same spatial section, the original linear position difference along the space can be converted into a three-dimensional position difference within the same section. Using the position that minimizes the sum of the Euclidean distances of each spatial intersection point to represent the building structure position can reduce the direct pull on the building structure position when the spatial intersection point of a single scanning station deviates from other scanning stations.

[0085] Furthermore, the actual locations of the building structures are arranged in chronological order according to the overall extension direction, and numbered sequentially starting from one end of the overall extension direction; when at least two different building structure locations are formed, the arithmetic mean of the three-dimensional coordinates of each building structure location is calculated to form the average location of the building structure locations, represented as: ; In the formula, This is the three-dimensional coordinate vector of the average position of the building structure. These are the arranged building structure location numbers; The actual number of building structures located; For the first The three-dimensional coordinate vector of each building structure location.

[0086] Furthermore, the difference vectors between the three-dimensional coordinates of each building structure location and the average position of each building structure location are generated. Each difference vector is multiplied by its own transpose, and then the arithmetic mean of all product matrices is calculated element-wise to form the position covariance matrix of the building structure locations, represented as: ; In the formula, This is the location covariance matrix of the building structure.

[0087] Furthermore, eigenvalue decomposition is performed on the location covariance matrix to obtain the eigenvector corresponding to the largest eigenvalue. The obtained eigenvector is then normalized to obtain the unit direction vector of the final building structure intersection line. Using the average position of the building structure locations as the base point of the line, the directed distance calculated from the average position along the unit direction vector of the final building structure intersection line is used as the position parameter of the final building structure intersection line. A spatial straight line passing through the average position and extending along the unit direction vector of the final building structure intersection line is established, represented as: ; In the formula, The final building structure intersection line is determined by positional parameters. Represents the three-dimensional coordinate vector of a point in space; The unit direction vector of the final building structure intersection line; It is the directed distance calculated from the average position along the unit direction vector of the final building structure intersection line, with dimensions consistent with three-dimensional coordinates.

[0088] The spatial straight line formed by extending the unit direction vector through the average position along the final building structure intersection line is taken as the final building structure intersection line of the two adjacent building surfaces.

[0089] When there are fewer than two actual building structure locations, or when all building structure locations are located in the same spatial location, the final building structure intersection line is not formed based on the current multi-scan site results.

[0090] It should be noted that due to differences in scanning distance, scanning direction, and building occlusion, the observation results of the same building structure intersection location at different scanning stations may have spatial offsets. If the intersection lines of building structures formed by different scanning stations are directly stitched together, or if all point clouds are mixed and then fitted to the intersection lines, the local measurement deviation generated by a single scanning station may be transmitted to the final intersection line, causing the final intersection line to deviate from the actual building connection location. Setting up spatial sections along the common extension direction of the intersection lines of building structures at each scanning station allows the observation results of different scanning stations at the same building structure location to be converted to the same extension location for comparison. Fitting the final intersection line of building structures based on the building structure location formed by multiple scanning stations within each spatial section ensures that the final intersection line is constrained by the observation results of multiple independent scanning stations, reducing the impact of the observation deviation of a single scanning station on the formation of the intersection line.

[0091] S4. Using the final building structure intersection line as the common geometric boundary of adjacent building surfaces, refit the building surfaces and construct a three-dimensional building model based on the refitted building surfaces.

[0092] Furthermore, the adjacent building surfaces that form the final building structure intersection line are processed group by group. The positions of each building structure arranged along the overall extension direction are read. The first and last building structure positions are orthogonally projected onto the final building structure intersection line, and the final building structure intersection line between the two projection points is used as the intersection line constraint range of the current adjacent building surfaces.

[0093] It should be noted that the final building structure intersection line is obtained by fitting multiple building structure positions formed along the overall extension direction. It can mathematically continue to extend to both ends, but only within the building structure range defined by the first and last building structure positions are the building structure positions jointly participated in by multiple scanning stations. Therefore, by projecting the first and last building structure positions onto the final building structure intersection line respectively, and only using the final building structure intersection line portion between the two projection points to refit the adjacent building surfaces, it is possible to both use the building structure positions jointly recovered by multiple scanning stations to correct the building surface connection boundaries, and avoid using the extended portion of the intersection line beyond the coverage of the building structure positions as the actual building surface connection boundaries.

[0094] Furthermore, the point clouds of the building surfaces belonging to two adjacent building surfaces are read separately; for the current building surface, the midpoint of the final building structure intersection between the two projection points is taken as the reference point for plane fitting, and the unit direction vector of the final building structure intersection is read; a candidate plane passing through the plane fitting reference point is established, the normal direction of the candidate plane is perpendicular to the unit direction vector of the final building structure intersection, so that the candidate plane contains the final building structure intersection; among the candidate planes, the candidate plane with the smallest sum of squared orthogonal distances from each point cloud in the building surface to the candidate plane is selected, denoted as: ; In the formula, This is the refitted unit normal vector of the building surface; Let be the unit normal vector of the candidate plane; The first in the current building surface The three-dimensional coordinate vector of a point cloud point; The three-dimensional coordinate vector of the reference point for plane fitting; Point cloud point number; This represents the number of point cloud points contained on the current building surface.

[0095] in, The candidate planes are defined to have their normal direction perpendicular to the unit direction vector of the intersection line of the final building structure. The candidate planes also pass through the plane fitting reference point located on the intersection line of the final building structure. Therefore, each candidate plane contains the intersection line of the final building structure. Among the candidate planes that simultaneously satisfy the perpendicular relationship of the normal and pass through the plane fitting reference point, the sum of the squares of the orthogonal distances of each point cloud point is compared. This allows us to obtain the building surface that is closest to the spatial distribution of the point cloud of the current building surface while keeping the position of the intersection line of the final building structure unchanged.

[0096] Specifically, in a two-dimensional directional space perpendicular to the unit direction vector of the final building structure intersection line, two mutually perpendicular unit direction vectors are selected. The spatial vectors of each point cloud point relative to the plane fitting reference point are projected onto the two unit direction vectors respectively, and a two-element column vector is formed according to the order of the two projection lengths. Each two-element column vector is multiplied by its own transpose and then accumulated to form a 2x2 scatter matrix. The 2x2 scatter matrix is ​​subjected to eigenvalue decomposition to obtain the eigenvector corresponding to the minimum eigenvalue. The eigenvector corresponding to the minimum eigenvalue represents the direction with the minimum dispersion of the building surface point cloud in the two-dimensional directional space. The two components of the eigenvector are used as the combination coefficients of the two unit direction vectors to form a spatial vector in a unified building coordinate system. The spatial vector is normalized to obtain the refitted building surface unit normal vector. A spatial plane passing through the plane fitting reference point and with the normal direction being the refitted building surface unit normal vector is established to form the refitted building surface.

[0097] Refit the building surface on the other side of the boundary in the same way, and calculate the inner product between the unit normal vector of the refitted building surface and the normal vector of the original building plane. When the inner product is less than zero, reverse the direction of the unit normal vector of the refitted building surface. When the inner product is not less than zero, keep the direction of the unit normal vector of the refitted building surface unchanged. Both refitted building surfaces contain the same final building structure intersection line, and the final building structure intersection line between the two projection points is taken as the common geometric boundary.

[0098] After refitting the two building surfaces, the point clouds of each building surface are orthogonally projected onto the refitted building surface. At the same time, the boundary points of the building surface boundaries extracted when forming the adjacency relationship are orthogonally projected onto the refitted building surface. The correspondence between each boundary point and the corresponding projected boundary point is recorded, and the projected boundary points are connected according to the original adjacency arrangement of each boundary point to form the boundary of the refitted building surface. The spatial sections located between the two projection points along the overall extension direction are read, and the spatial intersection points of the final building structure intersection line and each spatial section are calculated respectively. The first projection point, each spatial intersection point and the last projection point are arranged in sequence according to the overall extension direction. Points with overlapping spatial positions are removed, and the retained points are used as the common boundary vertices of the two building surfaces, and the arrangement order of each common boundary vertex is retained.

[0099] The continuous boundary portions used when the two building surfaces form an adjacent relationship are read separately. Based on the correspondence between each boundary point and the projected boundary points, the projected boundary points corresponding to the continuous boundary portions in the refitted building surface boundaries are obtained. Each projected boundary point is orthogonally projected onto the final building structure intersection line. When the obtained intersection line projection position is between the first projection point and the last projection point, the corresponding projected boundary point is removed from the refitted building surface boundary. According to the original arrangement direction of the current building surface boundary at the removed position, the common boundary vertices arranged along the overall extension direction are written into the boundary portion where the removed projected boundary points are located in either forward or reverse order. The projected boundary points outside the range between the first projection point and the last projection point maintain their original arrangement relationship, forming the updated building surface boundaries of the two building surfaces.

[0100] Furthermore, projection points within the updated building surface boundary but not coinciding with the building surface boundary in the refitted building surface are taken as internal vertices of the building surface. The updated building surface boundary is taken as the triangulation range. The line connecting two adjacent common boundary vertices is taken as constraint edges. Constrained Delaunay triangulation is performed on the two building surfaces respectively. Triangles located outside the building surface boundary and triangles with three vertices on the same straight line are deleted to form triangular meshes of the two building surfaces. The same common boundary vertex in the two building surfaces uses the same three-dimensional coordinates and vertex number, and the constraint edges between adjacent common boundary vertices are referenced together, so that the triangular meshes of the two building surfaces are continuously connected along the intersection line of the final building structure.

[0101] For adjacent building surfaces that have not yet formed the final building structure intersection line, triangular meshes are constructed using the already formed building planes and building surface boundaries; for curved roofs and architectural decorative curved surfaces that cannot be represented by planes, curved surface triangular meshes are reconstructed based on the retained building surface point cloud and the spatial adjacency relationships between the points in the point cloud.

[0102] The triangular meshes and curved triangular meshes of each building surface are collected, the mesh vertices that overlap in spatial position and are located at the connection position of the same building surface are merged, duplicate triangular facets with identical vertices are deleted, and the vertex arrangement order of the triangular facets is unified according to the original normal direction of each building surface to form a three-dimensional geometric model of the building.

[0103] Furthermore, the system reads architectural photographic images and their spatial calibration relationships with a unified architectural coordinate system. It then processes each triangular facet in the 3D architectural geometric model, transforming the three vertices of each facet to their pixel positions in the architectural photographs. When all three vertices are within the image range of the same architectural photograph, and the observation line segment from the camera position to the center of the facet does not intersect with any other facet before reaching the current facet, the current architectural photograph is used as a candidate texture source for the facet. When multiple candidate texture sources exist for the same facet, the smaller angle between the observation direction from the camera position to the center of the facet and the normal direction of the facet is obtained, and the architectural photograph with the smallest angle is selected. When the angles are the same, the architectural photograph with the smallest spatial distance from the camera position to the center of the facet is selected. Texture coordinates for the facet are established based on the pixel positions of the three vertices in the selected architectural photograph, and the image region enclosed by the three pixel positions is mapped to the facet. When no candidate texture source exists for a facet, its geometric information is preserved, and the corresponding unmapped image texture is recorded.

[0104] After completing the texture mapping of each triangular facet, the building point cloud, converted to a unified building coordinate system, is read. Based on the building surface affiliation of each point cloud point, the nearest triangular facet to the current point cloud point within the triangular mesh of the same building surface is found, and the spatial distance from the point cloud point to the nearest triangular facet is obtained. The average and maximum spatial distances of all point cloud points on each building surface are calculated to form the geometric accuracy check result of the 3D building model. Simultaneously, it is checked whether two adjacent building surfaces share a common boundary vertex with the same 3D coordinates and vertex number, and whether they jointly reference mesh edges between adjacent common boundary vertices, forming the building surface connectivity check result.

[0105] After completing the geometric accuracy check and the building surface connection check, the relationship between the building's three-dimensional geometric model, the texture coordinates of each triangular facet, and the building photographic images is recorded together. The results of the geometric accuracy check and the building surface connection check are then linked and recorded with the building's three-dimensional geometric model to form a building three-dimensional model that includes the three-dimensional geometric structure and building surface texture.

[0106] It should also be noted that when fitting adjacent building surfaces separately, even if each building surface can fit its own point cloud well, its boundary position may still not coincide due to the different number of point clouds, distribution density and viewing direction of different scanning stations. In this embodiment, the final building structure intersection line is introduced when refitting the building surface, so that the two adjacent building surfaces pass through the same building structure connection position, and the same common boundary vertex and mesh edge are used in the triangular mesh. The building surface point cloud participates in the fitting of the spatial position of the building surface, and the final building structure intersection line defines the connection boundary of adjacent building surfaces, thereby reducing the occurrence of boundary misalignment, model gaps, overlap or intersection at wall corners, ridges and the connection position between walls and roof.

[0107] It should also be noted that, in order to observe the construction of building structure intersections and the connection positions of adjacent building surfaces under multi-scanning-site conditions, a three-dimensional building model of the connection position between the wall and the sloping roof is established to construct a simulation scene. Different scanning sites have different scanning directions and actual scanning coverage areas. According to the present invention, building structure intersections and actual observation line segments are formed for each scanning site, and spatial sections are set along the overall extension direction. Only the spatial intersections of the scanning sites that cover the current spatial section are retained. Building structure positions are formed based on the spatial intersections within the same spatial section. Then, the final building structure intersections are fitted based on the building structure positions formed along the overall extension direction. Subsequently, the final building structure intersections are used as the common geometric boundaries of adjacent building surfaces to form common boundary vertices and constraint edges, and triangular meshes of adjacent building surfaces are constructed.

[0108] Figure 5This displays the number of scanning stations involved in forming the building structure's location and the changes in their spatial positions at different spatial cross-sectional locations. The horizontal axis represents the position of the spatial cross-section along the overall extension direction, the left vertical axis represents the spatial distance, and the right vertical axis represents the number of participating scanning stations. Solid dots represent the spatial distance from the building structure's location to the final intersection line, dashed triangles represent the median distance from the spatial intersection point to the building structure's location, and hollow columns represent the number of scanning stations involved in forming the building structure's location. Figure 5 It can be seen that the number of scanning stations involved in the formation of the building structure position changes in segments when moving along the overall extension direction, indicating that the actual number of scanning stations retained in different spatial sections is not the same; the distance from the spatial intersection point to the building structure position fluctuates to a certain extent, while the spatial distance from the building structure position to the final building structure intersection line remains within a small range overall; combined with the actual observation line segment's selection process for the spatial intersection point, it can be seen that the present invention can determine the segment involved in the formation of the building structure position according to the actual observation range of each scanning station, avoiding the mathematical extension of the building structure intersection line located outside the actual observation line segment from continuing to participate in the formation of the building structure position of the corresponding spatial section.

[0109] Figure 6 This diagram illustrates the triangular mesh connections between two adjacent building surfaces near the final structural intersection line. Thin solid lines represent the triangular mesh on the wall surface, thin dashed lines represent the triangular mesh on the sloping roof surface, gray dots represent the point cloud of the building surfaces, and thick black solid lines and solid dots represent the final structural intersection line and common boundary vertices. Figure 6 It can be seen that the triangular meshes of the wall building surface and the triangular meshes of the sloping roof building surface are connected along the same final building structure intersection line and pass through a continuous arrangement of common boundary vertices. The point cloud of the building surface has a certain discrete distribution near the corresponding triangular mesh, while the connection position of the two building surfaces remains continuous. It can be seen that by using the final building structure intersection line as the common geometric boundary and using common boundary vertices and constraint edges to construct the triangular meshes of the two building surfaces, adjacent building surfaces can adopt a unified connection position. From a geometric connection perspective, this avoids the inconsistency of connection positions when the two building surfaces form boundaries separately, and provides a basis for reducing boundary offsets, gaps, overlaps and intersections of the connection positions of the building 3D model.

[0110] In summary, this invention: by constraining the scanning stations with actual observed line segments to participate in the formation of the spatial cross-section building structure position, the building structure position is formed based on the spatial intersections within the same spatial cross-section, and the final building structure intersection line is fitted, thus distinguishing between the actual observed part and the mathematically extended part of the building structure intersection line and reducing the influence of the extended part of the intersection line on the building structure position; by using the final building structure intersection line as the common geometric boundary of adjacent building surfaces, and adopting common boundary vertices and constraint edges in the triangular mesh, the connection position of adjacent building surfaces is unified, reducing boundary offsets, gaps, overlaps, and intersections of the connection positions of the three-dimensional building model.

[0111] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for constructing a three-dimensional architectural model using laser scanning technology, characterized in that, include: The system acquires building point clouds at different scanning stations, performs coordinate transformation, registration, and building surface segmentation on the building point clouds to form adjacent building surfaces. Distance maps of building point clouds are recovered according to the scanning stations. Point clouds on both sides of the boundary are extracted based on the transformation position of the building surface and the local building plane is fitted to form the intersection line of the building structure and the actual observation line segment of each scanning station. Based on the overall extension direction formed by the intersection lines of each building structure, spatial sections are set along the overall extension direction. Scanning stations covering each spatial section are selected by actual observation line segments. The spatial intersection points of the building structure intersection lines and the spatial sections are obtained. The building structure positions are formed based on the spatial intersection points within the same spatial section. Finally, the building structure intersection lines are fitted based on the building structure positions formed along the overall extension direction. The building surfaces are refitted using the final building structure intersection line as the common geometric boundary of adjacent building surfaces, and a three-dimensional building model is constructed based on the refitted building surfaces.

2. The method for constructing a three-dimensional building model using laser scanning technology as described in claim 1, characterized in that, The acquisition of building point clouds at different scanning sites includes: Control points and multiple scanning stations are set up around the building. The building's facade, roof, wall corners, ridges, and the connection between the walls and the roof are scanned by a 3D laser scanner at each scanning station to form a building point cloud collected by different scanning stations. Each point cloud point records the 3D coordinates in the coordinate system of the scanning station, the original measurement distance, the scanning station identification, and the original scanning arrangement information.

3. The method for constructing a three-dimensional building model using laser scanning technology as described in claim 2, characterized in that, The coordinate transformation, registration, and surface segmentation of the building point cloud include: Based on the spatial correspondence of each control point in the scanning station coordinate system and the unified building coordinate system, a coordinate transformation relationship is formed to transform the building point clouds collected by different scanning stations to the unified building coordinate system, and the iterative nearest point algorithm is used to perform fine registration of the building point clouds in the overlapping scanning area. The building point cloud that has been finely registered is denoised and the local normal vector of each point cloud point is calculated. The region is segmented according to the spatial adjacency relationship between the point cloud points and the direction of the local normal vector to form the building surface point cloud. The building surface that can be represented by a plane is fitted with a plane to obtain the building plane and the plane equation. Project the building surface point cloud onto the building plane, extract the building surface boundary along the edge of the coverage area of ​​the projected point cloud in the building plane, and take the median value of the spatial distance between each point cloud point and the nearest point cloud point as the point distance between the building surface point cloud. The spatial intersection line is obtained based on the plane equations of two non-parallel building planes. Boundary points that are no more than the distance between the point cloud points on the respective building surfaces are extracted from the boundaries of the two building surfaces. Boundary points belonging to the same continuous boundary part are projected onto the spatial intersection line according to the adjacent arrangement of the boundary points to form projection line segments. When there are overlapping line segments in the projection line segments formed by the two building surfaces, the two building surfaces are regarded as adjacent building surfaces.

4. The method for constructing a three-dimensional building model using laser scanning technology as described in claim 3, characterized in that, The method of restoring the distance map of the building point cloud according to the scanning site includes: Adjacent building surfaces are processed in groups. The building point clouds collected by each scanning station are read according to the scanning station identifier, and the point cloud points are arranged into a two-dimensional scanning grid according to the original scanning arrangement information. The original measured distance of each point cloud point is written into the grid position where the point cloud point is located. At the same time, the three-dimensional coordinates of the point cloud point in the unified building coordinate system and the building surface to which it belongs are recorded at the grid position. The two-dimensional scanning grid that records the original measured distance is used as the distance map of the current scanning station. The positions of two directly adjacent grids that both record point cloud points are checked along the scan row and scan column directions of the distance map. When the two point cloud points belong to the two adjacent building surfaces being processed, the boundary between the two grid positions is recorded as the building surface transformation position. The building surface transformation positions with common endpoints are connected according to the scan row and column relationship to form a continuous transformation section.

5. The method for constructing a three-dimensional building model using laser scanning technology as described in claim 4, characterized in that, The step of extracting point clouds from both sides of the boundary based on the transformation position of the building surface and fitting the local building plane to form the intersection line of the building structure and the actual observed line segment of each scanning station includes: For each continuous transition section, the point cloud points belonging to two adjacent building surfaces on both sides of the transition position of the building surface are read, as well as the point cloud points that are directly adjacent to the grid position of the point cloud points and belong to the same building surface. The point cloud points are summarized according to the building surface to which they belong and duplicate point cloud points are removed to form the point cloud on both sides of the boundary. When the point clouds on both sides of the boundary each contain at least three non-collinear point cloud points, calculate the centroids of the point clouds on both sides of the boundary respectively. Form a covariance matrix based on the three-dimensional coordinate difference of each point cloud point relative to the centroid of the point cloud. Take the eigenvector corresponding to the smallest eigenvalue of the covariance matrix as the plane normal vector and establish local building planes that pass through the centroid of the point cloud and whose normal direction is the plane normal vector. When the two local building planes are not parallel, take the spatial intersection line of the two local building planes as the building structure intersection line of the current scanning station. The spatial midpoints of two point cloud points on both sides of the transformation position of each building surface are obtained respectively. The spatial midpoints are orthogonally projected onto the building structure intersection line of the current scanning station to form projection points. According to the arrangement order of each projection point along the building structure intersection line, when at least two projection points with different positions are formed in the same continuous transformation section, the part of the building structure intersection line between the first projection point and the last projection point is taken as the actual observation line segment, and the projection points in the actual observation line segment and the arrangement order of the projection points are retained.

6. The method for constructing a three-dimensional building model using laser scanning technology as described in claim 5, characterized in that, The overall extension direction formed by the intersection lines of each building structure includes: The adjacent building surfaces are processed in groups, and the intersection lines of the building structures and the actual observed line segments formed by different scanning stations are read. The unit direction vector of the intersection line of the building structures is formed based on the two endpoints of any actual observed line segment. Multiply the unit direction vector of the intersection line of the building structure at each scanning station with its own transpose vector to form a third-order direction matrix. Calculate the arithmetic mean of the third-order direction matrices formed by different scanning stations according to the matrix elements to form an average matrix. Perform eigenvalue decomposition on the average matrix and normalize the eigenvector corresponding to the largest eigenvalue to the unit direction vector of the overall extension direction.

7. The method for constructing a three-dimensional architectural model using laser scanning technology as described in claim 5 or 6, characterized in that, The process of setting up spatial cross-sections along the overall extension direction, selecting scanning stations that cover each spatial cross-section using actual observation line segments, and obtaining the spatial intersection points between the building structure intersection lines and the spatial cross-sections includes: The spatial distance between two adjacent projection points within the same actual observation line segment is obtained respectively. The median value of each spatial distance at the same scanning station is taken as the structural sampling interval. The median value of the structural sampling interval at different scanning stations is taken as the spatial cross-sectional interval distance. Select coordinate reference points from the endpoints of each actual observation line segment, and use the scalar projection of each actual observation line segment endpoint relative to the coordinate reference point along the overall extension direction as the position value of the endpoint; use the range between the minimum and maximum values ​​of the position values ​​of each endpoint as the setting range of the spatial section, and set the spatial sections sequentially according to the spatial section interval starting from the minimum position value, with the normal direction of each spatial section being consistent with the overall extension direction; Find the spatial intersection points of the building structure intersection lines and each spatial section at each scanning station. When the spatial intersection point is located between the two endpoints of any actual observation line segment at the current scanning station, the current scanning station is used as the scanning station whose actual observation line segment covers the current spatial section, and the spatial intersection point is retained.

8. The method for constructing a three-dimensional building model using laser scanning technology as described in claim 7, characterized in that, The process of forming the building structure position based on the spatial intersections within the same spatial cross-section, and fitting the final building structure intersection line based on the building structure positions formed along the overall extension direction, includes: When the same spatial section retains spatial intersections from no less than three different scanning stations, and the spatial intersections are not all located on the same straight line, the spatial position with the smallest sum of Euclidean distances of the spatial intersections shall be taken as the architectural position of the current spatial section. Arrange the locations of each building structure according to the overall extension direction. When at least two building structure locations with different positions are formed, calculate the arithmetic mean of the three-dimensional coordinates of each building structure location to form the average position of the building structure location. Form the difference vector between the three-dimensional coordinates of each building structure location and the average position of the building structure location. Multiply each difference vector with its own transpose vector and calculate the arithmetic mean of all product matrices by matrix elements to form the position covariance matrix of the building structure location. The location covariance matrix of the building structure is decomposed into eigenvalues. The eigenvector corresponding to the largest eigenvalue is normalized to the unit direction vector of the final building structure intersection line. The spatial straight line that passes through the average position of the building structure and extends along the unit direction vector of the final building structure intersection line is taken as the final building structure intersection line.

9. The method for constructing a three-dimensional architectural model using laser scanning technology as described in claim 8, characterized in that, The refitting of the building surface using the final building structure intersection line as the common geometric boundary of adjacent building surfaces includes: The first and last building structure positions arranged along the overall extension direction are orthogonally projected onto the final building structure intersection line, and the final building structure intersection line between the two projection points is used as the intersection line constraint range. Read the point clouds of the building surfaces of two adjacent building surfaces respectively, and take the midpoint of the space within the constraint range of the intersection line as the reference point for plane fitting; for each building surface, establish a candidate plane that passes through the reference point for plane fitting and whose normal direction is perpendicular to the unit direction vector of the intersection line of the final building structure; select the candidate plane with the smallest sum of squared orthogonal distances from each point cloud in the building surface to the candidate plane from each candidate plane, and form the refitted building surface containing the intersection line of the final building structure respectively. The intersection constraint range is used as the common geometric boundary of the two refitted building surfaces.

10. The method for constructing a three-dimensional architectural model using laser scanning technology as described in claim 9, characterized in that, The process of constructing a 3D building model based on the refitted building surface includes: The point clouds of the building surfaces of the two building surfaces are orthogonally projected onto the refitted building surfaces. The boundary points of the building surface boundaries extracted when forming the adjacency relationship are orthogonally projected onto the refitted building surfaces to form projected boundary points. The correspondence between each boundary point and the projected boundary points is preserved, and the projected boundary points are connected according to the original adjacent arrangement of each boundary point to form the refitted building surface boundary. Read each spatial section within the intersection constraint range, obtain the spatial intersection points of the final building structure intersection line and each spatial section respectively, arrange the projection points of the first and last building structure positions on the final building structure intersection line and each spatial intersection point according to the overall extension direction, remove the points with overlapping spatial positions, and form the common boundary vertex of the two building surfaces; Based on the correspondence between each boundary point and the projected boundary point, obtain the projected boundary points corresponding to the continuous boundary portion used when the two building surfaces form an adjacent relationship; orthogonally project the obtained projected boundary points to the final building structure intersection line, remove the projected boundary points whose intersection line projection positions are within the intersection line constraint range, and write the common boundary vertices into the boundary portion where the removed projected boundary points are located in the order of arrangement to form the updated building surface boundaries of the two building surfaces. Using the orthogonal projection positions of the building surface point cloud onto the refitted building surface as the internal vertices of the building surface, and the updated building surface boundary as the triangulation range, the lines connecting adjacent common boundary vertices are used as constraint edges. Constrained Delaunay triangulation is performed on the two building surfaces to form triangular meshes of the two building surfaces. The triangular meshes of the two building surfaces use common boundary vertices with the same 3D coordinates and vertex numbers, and jointly reference the constraint edges between adjacent common boundary vertices. Using the existing building plane and building surface boundaries, triangular meshes are constructed on adjacent building surfaces that have not yet formed the final building structure intersection line. Based on the retained building surface point cloud, curved triangular meshes are reconstructed on building surfaces that cannot be represented by a plane. The triangular meshes and curved triangular meshes of each building surface are summarized, and the mesh vertices that overlap in spatial position and are located at the connection position of the same building surface are merged. The vertex arrangement order of the triangular facets is unified according to the original normal direction of each building surface to form a three-dimensional building model.

Images