Building component instantiation extraction method based on indoor and outdoor point clouds and related equipment

Abstract

The present invention discloses a method for instantiating and extracting building components based on indoor and outdoor point clouds and related equipment. The method comprises: obtaining indoor and outdoor point clouds of a building for multi-scale point cluster segmentation, calculating point cluster plane index, extracting indoor structural plane point clusters and outdoor structural plane point clusters; extracting structural line segments, building floor heights and room corners of the building; analyzing the association relationship between components to obtain prior knowledge for identifying beams, columns and wall components in the building structure; based on the structural plane, using structural line segments, building floor heights and room corners as position references, and combining prior knowledge to perform position identification of components of different categories; generating beam component bounding boxes, column component bounding boxes and wall component bounding boxes based on component position identification symbols, and performing component point cloud segmentation on the building point cloud. The present invention effectively identifies beam, column and wall components, thereby quickly realizing component point cloud segmentation on the building point cloud.

Description

Technical Field

[0001] The present invention relates to the technical field of point cloud segmentation, and in particular to a method, system, terminal and computer-readable storage medium for instantiating and extracting building components based on indoor and outdoor point clouds. Background Art

[0002] In recent years, with the acceleration of urbanization and the rapid development of information and communication technology, the concept of digital twin cities has emerged. Digital twin cities require the digitization and real-time perception of various entities and elements in real cities. The three-dimensional reconstruction of urban buildings is being upgraded from the original object-level modeling for the purpose of digital visualization to component-level modeling for the purpose of intelligent application.

[0003] The traditional building component-level modeling method relies on professionals to determine the location of building components through field surveys or interpretation of building drawings, and then use professional software to perform three-dimensional modeling, which requires a lot of manpower and time costs during reconstruction. With the development of three-dimensional data acquisition technology, point clouds have become an ideal data carrier for urban buildings. Three-dimensional point clouds have high-precision geometric information and can accurately record building structural information. Accurately obtaining the locations of components of different categories based on fine-grained building identification based on three-dimensional point clouds is the key to achieving automated three-dimensional reconstruction of building component-level models. However, existing methods for automated building component identification based on three-dimensional point clouds mostly rely on the geometric or texture features of the target object features, which makes this type of method unable to be used to identify load-bearing components such as beams and columns embedded in the facade of the building.

[0004] Therefore, the prior art still needs to be improved and developed. Summary of the invention

[0005] The main purpose of the present invention is to provide a method, system, terminal and computer-readable storage medium for instantiating and extracting building components based on indoor and outdoor point clouds, aiming to solve the problem that the methods for automatically identifying building components based on three-dimensional point clouds in the prior art mostly rely on the geometric or texture features of the target object features, which makes this type of method unable to be used to identify load-bearing components such as beams and columns embedded in the facade of the building, i.e., leads to inaccurate identification of building components.

[0006] To achieve the above object, the present invention provides a method for instantiating and extracting building components based on indoor and outdoor point clouds, and the method for instantiating and extracting building components based on indoor and outdoor point clouds comprises the following steps:

[0007] Acquire indoor and outdoor point clouds of a building, perform multi-scale point cluster segmentation on the indoor and outdoor point clouds of the building, segment and separate different geometric structures in the building, calculate the point cluster plane index, and extract indoor structure plane point clusters and outdoor structure plane point clusters;

[0008] Extracting structural line segments of buildings based on the outdoor structural plane point clusters, and extracting building storey heights and room corner points based on the indoor structural plane point clusters;

[0009] Analyze the spatial and geometric relationships between components, and obtain prior knowledge for identifying beams, columns and wall components in building structures based on the spatial and geometric relationships.

[0010] Based on the outdoor structure plane point cluster and the indoor structure plane point cluster, taking the structure line segment, the building floor height and the room corner point as the position reference, and combining the prior knowledge, position identification of components of different categories is performed;

[0011] After obtaining the location of each category of components, the beam component bounding box, column component bounding box and wall component bounding box are generated based on the component position identification symbol, and the building point cloud is segmented into components.

[0012] In addition, to achieve the above-mentioned purpose, the present invention also provides a building component instantiation extraction system based on indoor and outdoor point clouds, wherein the building component instantiation extraction system based on indoor and outdoor point clouds comprises:

[0013] A multi-scale point cluster segmentation module is used to obtain indoor and outdoor point clouds of buildings, perform multi-scale point cluster segmentation on the indoor and outdoor point clouds of buildings, segment and separate different geometric structures in the building, calculate the point cluster plane index, and extract indoor structure plane point clusters and outdoor structure plane point clusters;

[0014] A geometric element extraction module, used to extract the structural line segments of the building based on the outdoor structural plane point clusters, and to extract the building floor heights and room corner points based on the indoor structural plane point clusters;

[0015] The correlation analysis module is used to analyze the spatial correlation and geometric correlation between components, and obtain the prior knowledge for identifying beams, columns and wall components in the building structure based on the spatial correlation and geometric correlation.

[0016] A component position identification module, which is used to identify the positions of components of different categories based on the outdoor structural plane point cluster and the indoor structural plane point cluster, the structural line segments, the building floor heights and the room corner points as position references, and in combination with the prior knowledge;

[0017] The component point cloud segmentation module is used to obtain the location of each category of components, generate beam component bounding boxes, column component bounding boxes and wall component bounding boxes based on the component position identification symbols, and perform component point cloud segmentation on the building point cloud.

[0018] In addition, to achieve the above-mentioned purpose, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a building component instantiation extraction program based on indoor and outdoor point clouds stored in the memory and executable on the processor, wherein the building component instantiation extraction program based on indoor and outdoor point clouds implements the steps of the building component instantiation extraction method based on indoor and outdoor point clouds as described above when executed by the processor.

[0019] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a building component instantiation extraction program based on indoor and outdoor point clouds, and when the building component instantiation extraction program based on indoor and outdoor point clouds is executed by a processor, the steps of the building component instantiation extraction method based on indoor and outdoor point clouds as described above are implemented.

[0020] In the present invention, indoor and outdoor point clouds of a building are obtained, and multi-scale point cluster segmentation is performed on the indoor and outdoor point clouds of the building. After different geometric structures in the building are segmented and separated, the point cluster plane index is calculated, and the indoor structure plane point cluster and the outdoor structure plane point cluster are extracted; the structural line segments of the building are extracted based on the outdoor structure plane point cluster, and the building floor height and room corner points are extracted based on the indoor structure plane point cluster; the spatial association relationship and the geometric association relationship between the components are analyzed, and according to the spatial association relationship and the geometric association relationship, the prior knowledge for identifying beams, columns and wall components in the building structure is obtained; based on the outdoor structure plane point cluster and the indoor structure plane point cluster, the position of components of different categories is identified with the structural line segments, the building floor height and the room corner points as position references, and combined with the prior knowledge; after obtaining the position of each category of components, the beam component bounding box, the column component bounding box and the wall component bounding box are generated based on the component position identification symbol, and the building point cloud is segmented into component point clouds. The present invention extracts geometric elements of different dimensions based on the indoor and outdoor point clouds of buildings, provides data support and accurate position reference for subsequent component identification, and summarizes simple prior knowledge for component identification under the condition of analyzing the spatial and geometric association relationships of components, thereby effectively identifying beams, columns and wall components, thereby quickly realizing component point cloud segmentation of the building point cloud. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of a preferred embodiment of the instantiation extraction method of building components based on indoor and outdoor point clouds of the present invention;

[0022] Figure 2 It is a flow chart of instantiated extraction of load-bearing components of a building in a preferred embodiment of the instantiated extraction method of building components based on indoor and outdoor point clouds of the present invention;

[0023] Figure 3 It is a schematic diagram of generating the bounding box of beam and column components in a preferred embodiment of the instantiation extraction method of building components based on indoor and outdoor point clouds of the present invention;

[0024] Figure 4 It is a schematic diagram of generating a wall component bounding box in a preferred embodiment of the instantiation extraction method of building components based on indoor and outdoor point clouds of the present invention;

[0025] Figure 5 It is a structural diagram of a preferred embodiment of the instantiation extraction system of building components based on indoor and outdoor point clouds of the present invention;

[0026] Figure 6 It is a structural diagram of a preferred embodiment of the terminal of the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present invention clearer and more specific, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0028] The instantiation extraction method of building components based on indoor and outdoor point clouds described in the preferred embodiment of the present invention is as follows: Figure 1 and Figure 2 As shown, the instantiation extraction method of building components based on indoor and outdoor point clouds includes the following steps:

[0029] Step S10, obtaining indoor and outdoor point clouds of the building, performing multi-scale point cluster segmentation on the indoor and outdoor point clouds of the building, segmenting and separating different geometric structures in the building, calculating the point cluster plane index, and extracting indoor structure plane point clusters and outdoor structure plane point clusters.

[0030] Specifically, an outdoor point cloud of a building is obtained, supervoxel segmentation is performed on the outdoor point cloud of the building, adjacent supervoxels are merged using a region growing algorithm, and a set of outdoor multi-scale point clusters with plane features preserved is generated, and outdoor plane structure point clusters are extracted from the outdoor multi-scale point cluster set according to the calculated outdoor point cluster plane index; an indoor point cloud of a building is obtained, supervoxel segmentation is performed on the indoor point cloud of the building, adjacent supervoxels are merged using a region growing algorithm, and a set of indoor multi-scale point clusters with plane features preserved is generated, and indoor plane structure point clusters are extracted from the indoor multi-scale point cluster set according to the calculated indoor point cluster plane index.

[0031] The outdoor point cloud of the building is segmented by supervoxel using a heuristic supervoxel segmentation algorithm based on local information. Supervoxel refers to a set of points with a certain homogeneity. Compared with discrete points, supervoxel has the characteristics of preserving local planar geometric features. In order to give full play to the characteristics of supervoxel feature preservation and boundary preservation, a heuristic supervoxel segmentation algorithm (Boundary Preserved Supervoxel Segmentation, BPSS) based on local information is adopted. The algorithm uses an energy function to describe the supervoxel set. The energy function used by the heuristic supervoxel segmentation algorithm consists of a geometric difference term and a point cluster quantity term:

[0032]

[0033] Among them, E(Z) represents the entire energy function, z ij To determine the point p j Does it belong to the point p i In the supervoxel represented by, λ is the regularization coefficient, which is used to weigh the weight of the geometric difference term and the number of point clusters in the energy function, C(Z) represents the number of supervoxels currently generated, K represents the expected number of supervoxels obtained according to the preset resolution, and D(p i ,p j ) is a similarity measurement parameter used to measure the similarity between two points, N represents the number of point clusters, min represents minimization, and st represents subject to, which means subject to.

[0034] Among them, the similarity measurement parameter D(p i ,p j ) is defined as:

[0035]

[0036] Among them, D s is the Euclidean distance between two points, D n is the normal vector n of the two points i and n j The dot product of R is the predefined supervoxel resolution, and to adapt to the density difference of different point clouds, and to ensure that the supervoxel can effectively maintain the local geometric structure characteristics of the ground object, according to the results of multiple experiments, R is set to 8 times the original point cloud density, μ and ∈ are D s and D n To ensure the local planar characteristics of the supervoxel, we set μ = 0.4, ∈ = 1.

[0037] By solving the optimal solution of the above energy function, supervoxel segmentation of three-dimensional point cloud data (outdoor point cloud of buildings) is achieved. The specific algorithm flow is as follows:

[0038] (1) Assuming that there are N points in the point cloud, a three-dimensional point cloud spatial index is established based on KD-Tree (k-dimensional tree, tree data structure) to search for point p i The k neighboring points of constitute the neighborhood point set P N , based on P N Calculate the normal vector for each point Repeat the above steps for each point in the point cloud.

[0039] (2) Divide the point cloud into voxels according to a preset supervoxel resolution R, and set the expected number of supervoxels K to the number of supervoxels to be segmented.

[0040] (3) Calculate the minimum similarity metric parameter D between each point in the point cloud and its domain point min , forming the original similarity measurement parameter set Let the initial value of λ be the median of D.

[0041] (4) Assume that each point represents a supervoxel S i = {r i ,G i}, forming the original supervoxel set S = {S1,…,S N}, where r i Supervoxel S i The representative point, G i is a supervoxel subset, whose initialization value is r i The neighborhood set of .

[0042] (5) Take p j For r i A neighborhood point of i =1, if λ-c i D(r i ,p j )<0, then p j The subset of supervoxels represented by G j Merge into G i , and p j The supervoxel S j Remove from S.

[0043] (6) Repeat steps (3) and (4) for each point in the point cloud. In each cycle, λ is doubled until the size of the supervoxel set S is equal to the estimated expected number of supervoxels K, completing the original supervoxel segmentation.

[0044] (7) In order to make the supervoxel better preserve the boundary information, the energy function is further optimized, and p is taken j For p i A neighboring point of j and pi do not belong to the same supervoxel, then D(p i ,r j ) is less than D(p i ,r i ), if it is less than p i Add to p j In the supervoxel.

[0045] (8) Repeat step (7) for each point in the point cloud to complete the boundary optimization of the original supervoxel.

[0046] Then, the region growing algorithm is used to merge adjacent supervoxels. The basic algorithm process is as follows: first, supervoxels are selected as seed supervoxels in the growth order; then, the supervoxels are grown according to the neighborhood relationship between supervoxels, and only supervoxels with planar similarity are allowed to be segmented into the same point cluster; once no new supervoxel is segmented into the point cluster where the seed supervoxel is located, the next seed supervoxel is selected for growth until all supervoxels are grown; when all planar point clusters are segmented, the unmerged supervoxels are merged with the planar point cluster set to generate an outdoor multi-scale point cluster set with planar feature preservation.

[0047] In order to extract the plane structure point cluster from the multi-scale point cluster set, let a multi-scale point cluster be C, and fit the multi-scale point cluster C into a plane equation based on the least squares method:

[0048] Π(x, y, z): Ax+By+C'z+D =0; (4)

[0049] Among them, Π represents the plane structure, A, B, C' are the three coefficients of the plane equation, and D is a constant.

[0050] The residual r is obtained by calculating the average distance from the midpoint of the multi-scale point cluster C to the fitting plane:

[0051]

[0052] Where n represents the number of multi-scale point clusters, and i represents any one of the multi-scale point clusters;

[0053] In addition, suppose p i (x i ,y i , z i ) point in the cluster, obtain the eigenvalue to calculate the plane feature of the cluster, perform eigendecomposition on the covariance matrix Q of the multi-scale cluster C to obtain the eigenvalues ​​λ1, λ2, λ3 of the cluster, where λ1>λ2>λ3, and then use formula (6) to obtain the plane feature p of the cluster Π :

[0054]

[0055] Define the point cluster size c as the number of points contained in the point cluster, and set the normalized point cluster size to:

[0056]

[0057] Among them, c max is the point cluster size of the largest point cluster in the outdoor multi-scale point cluster set.

[0058] In order to fully evaluate the plane structure characteristics of the point cluster, the point cluster plane index P is calculated according to formula (8): Π :

[0059]

[0060] Finally, the threshold t is set according to the point cluster plane index P Π And threshold t extracts outdoor plane structure point clusters from the outdoor multi-scale point cluster set. If the point cluster plane index P Π If it is greater than the threshold t, it belongs to the outdoor plane structure point cluster.

[0061] In addition, the process of extracting indoor plane structure point clusters is consistent with the above-mentioned process of extracting outdoor plane structure point clusters.

[0062] Step S20: extracting structural line segments of the building based on the outdoor structural plane point clusters, and extracting building floor heights and room corner points based on the indoor structural plane point clusters.

[0063] Specifically, edge line extraction technology is used to extract edge lines of the outdoor plane structure point cluster to obtain multiple three-dimensional line segments. There are many noise line segments and useless line segments that do not contain building component position information in the three-dimensional line segments extracted by the above method. Therefore, the extracted original line segments are post-processed. The specific process is as follows:

[0064] According to the characteristics of parallelism and orthogonality of building structures, it is concluded that the line segments based on the building structure should also have parallel and orthogonal relationships. Based on the above idea, multiple three-dimensional line segments are arranged in descending order according to their lengths, and multiple three-dimensional line segments are clustered according to their relative directions. The three-dimensional line segment with the longest length is taken as the starting reference line segment L of the first cluster lc1; the line segments to be clustered l are calculated in sequence according to the order of line segment arrangement. i and reference line segment l r The angle between like It is believed that l i ∈lc1, if Then create a new cluster lc i , think l i is the clustering lc ireference line segments; repeat the above operation until all three-dimensional line segments are clustered. For the three-dimensional line segments that are not clustered, they are determined to be noise line segments and are removed; after the line segment clustering is completed, the sum of the line segment lengths in each cluster is calculated, and the line segment clusters with the first two digits of the sum of the line segment lengths are set as the potential structure line segment set of the outdoor plane structure point cluster.

[0065] The above method can effectively remove the unstructured noise point clusters, but the line segment set also includes regular contour line segments of doors and windows. According to the building construction rules, the length and width of doors and windows are much smaller than the length and width of the facades where they are located. Therefore, the line segments without component information are removed according to the line segment length. At the same time, in order to make the line segments retain the building component information to the greatest extent, the line segments without component information are removed according to the line segment length. According to the structural index St of the line segment cluster, the line segment with no component information is removed. Π Dynamically adjust the rejection threshold th (pre-set the rejection threshold th); among them, the structural index St ∏ The calculation is: calculate the angle between the line segments in the line segment set two by two like || means or, then the line segment l is considered i and l i+1 It has structurality, and the ratio of the sum of the lengths of the structural line segments to the sum of the lengths of all line segments in the line segment set is calculated as the structural index St of the outdoor plane structure point cluster. Π .

[0066] Assume that the length of the longest line segment in the outdoor plane structure point cluster is L', and dynamically adjust the elimination threshold th:

[0067]

[0068] If the length of the line segment within the outdoor plane structure point cluster is less than th, it will be eliminated.

[0069] Finally, calculate the remaining line segments l and the perpendicular line l in the structure line cluster ⊥ Angle of direction (0,0,1) like The line segment l is considered to be a vertical structural line segment and is retained to obtain the structural line segment of the building.

[0070] The present invention extracts room corner points after spatially segmenting the indoor structural plane, and provides a position reference for the subsequent identification and extraction of components in the overall structure of the building. The present invention first uses the above method to extract the indoor plane structure point clusters from the indoor point cloud of the building. Considering that the indoor space is a closed area surrounded by vertical facades, the horizontal structure point clusters and the vertical structure point clusters are first distinguished according to the normal vector direction of the structure plane point clusters, and the ground bottom plate and top plate point clusters are extracted according to the elevation of the horizontal point clusters. The building floor height f is calculated according to the average distance between the floor point clusters and the top plate point clusters, and the vertical point clusters are projected to the plane where the ground point clusters are located to generate a binary mapping of the indoor space; the morphological segmentation method and the regular grid division method are combined on the binary mapping to extract the two-dimensional room corner points; finally, the two-dimensional room corner points are mapped back to the three-dimensional space, and the inaccurate extraction of the room corner points caused by the shared wall problem is optimized. The specific optimization process is as follows:

[0071] Take a room corner point p, and search for neighboring points of the corner point p with the thickness of the component t' (according to the common component thickness, the present invention sets t' = 0.15 meters) as the neighborhood search radius, and construct a neighborhood point set N{p, p2, p3, ..., p n}, calculate the center point p of the point set c As the final room corner point.

[0072] Step S30: Analyze the spatial association relationship and geometric association relationship between components, and obtain prior knowledge for identifying beams, columns and wall components in the building structure based on the spatial association relationship and geometric association relationship.

[0073] Specifically, the spatial association relationship includes: adjacent relationship, nested relationship, supporting relationship, connecting relationship and hanging relationship; the present invention considers that there are the following five kinds of spatial association relationships between building components:

[0074] (1) Adjacent relationship: The existence of contact surfaces between components is a typical manifestation of adjacent relationship. For example, the side of a wall usually contacts the side of a column.

[0075] (2) Nested relationship: One component completely contains another component, for example, columns and beams are embedded in the wall of a building.

[0076] (3) Support relationship: One component bears the weight of another component. For example, the column bears the building load at a vertical angle, and the beam bears the load brought by the building top or bottom plate at a horizontal angle.

[0077] (4) Connection relationship: Components are connected to each other through connectors, for example, columns and beams are connected through steel connectors or concrete connectors.

[0078] (5) Suspension relationship: One component is suspended on another component, for example, a beam is usually suspended on a column.

[0079] The geometric association relationship is a geometric abstraction of the spatial association relationship, including: parallel relationship, vertical relationship, alignment relationship, tangent relationship and separation relationship; the geometric association relationship of components is a geometric abstraction of the spatial association relationship, mainly including the following five types:

[0080] (1) Parallel relationship: Two components are parallel and non-repetitive on the same plane. For example, two column components in the same plane structure are usually parallel.

[0081] (2) Vertical relationship: Two components are perpendicular and do not overlap. For example, the beam and column components in the same plane structure are usually perpendicular.

[0082] (3) Alignment relationship: Components are arranged in the same direction in the same plane. For example, column components in the same plane structure are aligned in the horizontal direction, and beam components are aligned in the vertical direction.

[0083] (4) Tangent relationship: Two components are tangent in a plane. For example, beams, columns, and wall components are interconnected in actual space, which is geometrically manifested as different faces of the component geometry being tangent.

[0084] (5) Separation relationship: The components are not tangent in space. For example, components in different plane structures are usually separated.

[0085] Based on the above-mentioned spatial association relationship and geometric association relationship, the following prior knowledge for identifying beams, columns and strong components in building structures can be obtained, that is, the prior knowledge includes: first prior knowledge, second prior knowledge, third prior knowledge and fourth prior knowledge.

[0086] (1) The first prior knowledge is that there are column components at the intersection of different structural walls.

[0087] (2) The second prior knowledge is that columns and walls are perpendicular to the ground, and beams are horizontal to the ground.

[0088] (3) The third prior knowledge is that the top plate, the bottom plate are connected to the beams and are suspended on the columns.

[0089] (4) The fourth prior knowledge is that the wall, beam and column are interconnected and located in the same plane structure.

[0090] Step S40: Based on the outdoor structure plane point cluster and the indoor structure plane point cluster, taking the structure line segment, the building floor height and the room corner point as position reference, and combining the prior knowledge, position identification of components of different categories is performed.

[0091] Specifically, the position of the column components is first identified. According to the first prior knowledge and the second prior knowledge, it is concluded that there are column components at the structural line segments extracted based on the plane structure of the building's outdoor point cloud, but only two column identification line segments can be extracted from a plane structure Π. The column components extracted by the above method are not complete. Therefore, it is further inferred based on the first prior knowledge that there are column components at the corners of the indoor rooms of the building. In order to accurately obtain the position of the column components at the corners of the building on the building's facade, a corner point set P'{p'1,p'2…,p' n’}, the points in the corner point set P' are sorted in clockwise order, n' represents the number of corner points; let p' i is a point in P', according to p' i Construct two straight lines from the left and right corners of (If i=1, then i-1=n'), (If i = n', then i + 1 = 0) Find the straight line The intersection points with the plane structure Π of the building facade create an intersection point set Q{q1,q2…,q n”}, n" represents the number of intersection points; calculate p' i With the intersection point q i If the distance d is greater than the thickness t' of the component, then q i Remove from the intersection set Q, the remaining intersection points Q{q1,q2…,q m} represents the corner point p' i For the position of the column component on the plane structure Π, m represents the number of remaining intersections; in order to further accurately identify the column component, a column identification line segment l is created with the intersection as the starting point according to the direction and length of the vertical structural line on the plane. C .

[0092] After the position of the column component is identified, the present invention identifies the beam component according to the third prior knowledge and the fourth prior knowledge; assuming that the beam identification line segment l B It is the position mark of the beam component on the plane structure. The specific steps are as follows:

[0093] (1) Assume that there is a set of original column identification line segments on the plane structure Π of the building. The line segments in the set are arranged in a clockwise direction on the plane.

[0094] (2) Take the column identification line and in, They are the starting points of the two marking lines close to the ground. The endpoint of the other end of the marking line.

[0095] (3) Derive the distance between the two starting points along the vertical direction of the marking line to obtain beam B i The two end nodes and The beam identification line

[0096] (4) Node-based and Column identification line and Split and generate new column identification lines as well as And update the original identification line, after updating, it becomes and

[0097] (5) Iterate steps (3) and (4) until and The length is less than the building floor height f.

[0098] (6) Repeat the above steps until every column identification line segment is involved in the calculation.

[0099] The wall component is a horizontal plane with thickness in the actual three-dimensional space, and its position cannot be marked by an identification line. Therefore, the present invention uses a horizontal plane composed of four nodes. W (n1, n2, n3, n4) identifies the location of the wall component. After determining the locations of the column components and the beam components on the plane structure of the building, it is known based on the fourth prior knowledge that the nodes of the wall components coincide with the intersections of the beam and column components. Since the present invention divides the original column identification lines according to the positions of the beam components during the process of identifying the beam components, the nodes of the wall components are the first and last endpoints of two adjacent column identification lines after division.

[0100] Step S50: After obtaining the location of each category of components, a beam component bounding box, a column component bounding box and a wall component bounding box are generated based on the component position identification symbol, and the building point cloud is segmented into component point clouds.

[0101] Specifically, after obtaining the location of each component, the present invention generates a corresponding component bounding box based on the component location identifier to segment the outdoor point cloud of the building (for example, the building point cloud before and after damage).

[0102] (1) The same component bounding box generation strategy is used for beam and column components. The specific steps are as follows:

[0103] 1) Move the endpoint of the identified line segment along the normal vector n of the plane structure Π Π Positive and negative direction derivatives The distance of the component bounding box part vertex v′(v1,v2,v3,v4) is obtained, such as Figure 3 As shown in (a) in .

[0104] 2) Vertices n1, n2, n3, and n4 form a horizontal plane Π'. Let the normal vector of the horizontal plane Π' be n Π′ .

[0105] 3) Then mark the endpoint of the line segment along the normal vector n Π′ Positive and negative direction derivatives The distance to get the remaining vertices v″(v5,v6,v7,v8), such as Figure 3 As shown in (b) in .

[0106] 4) Merge the vertices v′(v1, v2, v3, v4) of the component bounding box and the remaining vertices v″(v5, v6, v7, v8) to obtain the component bounding box vertices V(v1, v2, v3, v4, v5, v6, v7, v8);

[0107] 5) Finally, connect the vertices in the component bounding box V (v1, v2, v3, v4, v5, v6, v7, v8) according to the preset connection method (such as Figure 3 As shown in (c) in the figure, they are connected to form the beam component enclosing box and the column component enclosing box.

[0108] Compared with the position identification lines obtained after beam and column identification, the wall components can obtain a more specific location range after identification, so the component bounding box generation method is more convenient. Figure 4 First, let the wall component identification surface Π W The normal vector of (n1,n2,n3,n4) (i.e. the horizontal plane formed by the four nodes) is Then derive each node of the identified surface along the positive and negative directions of the normal vector Get the vertex set of the component bounding box by Finally, connect the vertices on the same horizontal plane and then connect the corresponding positive and negative direction nodes in sequence to generate the wall component bounding box.

[0109] After completing the generation of each component bounding box (beam component bounding box, column component bounding box and wall component bounding box), the building point cloud is segmented by combining the theoretical knowledge of spatial geometry. By comparing the midpoint p' of the point cloud with the center point p' of the component bounding box SSIBB c To the different planes of the component bounding box i’ The difference in distance d i’ , the component bounding box includes a beam component bounding box, a column component bounding box and a wall component bounding box, and the building point cloud is segmented into component point clouds:

[0110] d i’ =dis(plane i’ ,p' c )-dis(plane i’ ,p”)st{i'∈Z|1≤i'≤6}; (10)

[0112] Among them, dis(plane i’ ,p”) represents the distance from point p” in the point cloud to the i’th surface of the component, dis(plane i’ ,p' c ) represents the center point p' of the component bounding box SSIBB c The distance to the i'th face of the component, Z represents the number of faces of the constructed bounding box.

[0113] If all d in {i'∈Z|1≤i'≤6} i’ If both are less than 0, the current point is considered to belong to the component point cloud.

[0114] The present invention firstly divides the outdoor point cloud of the building from the geometric morphology level through a multi-scale point cluster segmentation method, then extracts the plane structure point cluster from it by calculating the planarity, residual, size and other features of the multi-scale point cluster, and then extracts the structural line segment based on the plane structure point cluster while taking into account the characteristics of the building structure, and also extracts the indoor corner point using the indoor point cloud of the building. The present invention extracts geometric elements of different dimensions based on the indoor and outdoor point clouds of the building, providing data support and accurate position reference for subsequent component identification.

[0115] The present invention summarizes simple prior knowledge for component identification under the condition of analyzing the spatial and geometric association relationship of components. Based on these rules, the present invention first proposes a method for identifying the position of column components based on structural line segments and room corners. Furthermore, the present invention identifies beam components based on the position of column components and the mutual association relationship between beam and column components combined with the building height information. Finally, the position of wall components is identified based on the beam-column identification structure.

[0116] Furthermore, if Figure 5 As shown, based on the above-mentioned building component instantiation extraction method based on indoor and outdoor point clouds, the present invention also provides a building component instantiation extraction system based on indoor and outdoor point clouds, wherein the building component instantiation extraction system based on indoor and outdoor point clouds includes:

[0117] The multi-scale point cluster segmentation module 51 is used to obtain the indoor and outdoor point clouds of the building, perform multi-scale point cluster segmentation on the indoor and outdoor point clouds of the building, segment and separate different geometric structures in the building, calculate the point cluster plane index, and extract the indoor structure plane point cluster and the outdoor structure plane point cluster;

[0118] A geometric element extraction module 52, configured to extract structural line segments of a building based on the outdoor structural plane point clusters, and extract building floor heights and room corner points based on the indoor structural plane point clusters;

[0119] The correlation analysis module 53 is used to analyze the spatial correlation and geometric correlation between the components, and obtain the prior knowledge for identifying the beam, column and wall components in the building structure according to the spatial correlation and geometric correlation;

[0120] A component position identification module 54 is used to identify the positions of components of different categories based on the outdoor structural plane point cluster and the indoor structural plane point cluster, the structural line segment, the building floor height and the room corner point as position references, and in combination with the prior knowledge;

[0121] The component point cloud segmentation module 55 is used to obtain the location of each category of components, generate beam component bounding boxes, column component bounding boxes and wall component bounding boxes based on component position identification symbols, and perform component point cloud segmentation on the building point cloud.

[0122] Furthermore, if Figure 6 As shown, based on the above-mentioned method and system for instantiating and extracting building components based on indoor and outdoor point clouds, the present invention also provides a terminal accordingly, and the terminal includes a processor 10, a memory 20 and a display 30. Figure 6 Only some components of the terminal are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0123] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory of the terminal. In other embodiments, the memory 20 may also be an external storage device of the terminal, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the terminal. Further, the memory 20 may also include both an internal storage unit of the terminal and an external storage device. The memory 20 is used to store application software and various types of data installed in the terminal, such as the program code of the installation terminal. The memory 20 may also be used to temporarily store data that has been output or is to be output. In one embodiment, the memory 20 stores a building component instantiation extraction program 40 based on indoor and outdoor point clouds, and the building component instantiation extraction program 40 based on indoor and outdoor point clouds can be executed by the processor 10, thereby realizing the building component instantiation extraction method based on indoor and outdoor point clouds in the present application.

[0124] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run the program code or process data stored in the memory 20, such as executing the instantiation extraction method of building components based on indoor and outdoor point clouds.

[0125] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 30 is used to display information on the terminal and to display a visual user interface. The components 10-30 of the terminal communicate with each other via a system bus.

[0126] In one embodiment, when the processor 10 executes the building component instantiation extraction program 40 based on indoor and outdoor point clouds in the memory 20, the steps of the building component instantiation extraction method based on indoor and outdoor point clouds as described above are implemented.

[0127] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a building component instantiation extraction program based on indoor and outdoor point clouds, and when the building component instantiation extraction program based on indoor and outdoor point clouds is executed by a processor, the steps of the building component instantiation extraction method based on indoor and outdoor point clouds as described above are implemented.

[0128] In summary, the present invention provides a method, system, terminal and storage medium for instantiating and extracting building components based on indoor and outdoor point clouds. The method includes: obtaining indoor and outdoor point clouds of the building, performing multi-scale point cluster segmentation on the indoor and outdoor point clouds of the building, segmenting and separating different geometric structures in the building, calculating the point cluster plane index, extracting indoor structure plane point clusters and outdoor structure plane point clusters; extracting the structural line segments of the building based on the outdoor structure plane point clusters, and extracting the building floor height and room corner points based on the indoor structure plane point clusters; analyzing the spatial association relationship and geometric association relationship between components, and obtaining prior knowledge for identifying beams, columns and wall components in the building structure according to the spatial association relationship and the geometric association relationship; based on the outdoor structure plane point clusters and the indoor structure plane point clusters, taking the structural line segments, the building floor height and the room corner points as position references, and combining the prior knowledge to perform position identification of components of different categories; after obtaining the position of each category of components, generating beam component bounding boxes, column component bounding boxes and wall component bounding boxes based on component position identification symbols, and performing component point cloud segmentation on the building point cloud. The present invention extracts geometric elements of different dimensions based on the indoor and outdoor point clouds of buildings, provides data support and accurate position reference for subsequent component identification, and summarizes simple prior knowledge for component identification under the condition of analyzing the spatial and geometric association relationships of components, thereby effectively identifying beams, columns and wall components, thereby quickly realizing component point cloud segmentation of the building point cloud.

[0129] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or terminal including the element.

[0130] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the program can be stored in a computer-readable storage medium that can be read by a computer, and the program can include the processes of the above-mentioned method embodiments when executed. The computer-readable storage medium can be a memory, a disk, an optical disk, etc.

[0131] It should be understood that the application of the present invention is not limited to the above examples. For ordinary technicians in this field, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A building component instantiation extraction method based on indoor and outdoor point clouds, characterized in that: The instantiation extraction method of building components based on indoor and outdoor point clouds includes: Acquire indoor and outdoor point clouds of a building, perform multi-scale point cluster segmentation on the indoor and outdoor point clouds of the building, segment and separate different geometric structures in the building, calculate the point cluster plane index, and extract indoor structure plane point clusters and outdoor structure plane point clusters; Extracting structural line segments of buildings based on the outdoor structural plane point clusters, and extracting building storey heights and room corner points based on the indoor structural plane point clusters; Analyze the spatial and geometric relationships between components, and obtain prior knowledge for identifying beams, columns and wall components in building structures based on the spatial and geometric relationships. Based on the outdoor structure plane point cluster and the indoor structure plane point cluster, taking the structure line segment, the building floor height and the room corner point as the position reference, and combining the prior knowledge, position identification of components of different categories is performed; After obtaining the location of each category of components, the beam component bounding box, column component bounding box and wall component bounding box are generated based on the component position identification symbol, and the building point cloud is segmented into component point clouds; The method of obtaining indoor and outdoor point clouds of a building, performing multi-scale point cluster segmentation on the indoor and outdoor point clouds of the building, segmenting and separating different geometric structures in the building, calculating the point cluster plane index, and extracting indoor structure plane point clusters and outdoor structure plane point clusters specifically includes: Acquire an outdoor point cloud of a building, perform supervoxel segmentation on the outdoor point cloud of the building, merge adjacent supervoxels using a region growing algorithm, and generate an outdoor multi-scale point cluster set with plane feature preservation, and extract an outdoor plane structure point cluster from the outdoor multi-scale point cluster set according to the calculated outdoor point cluster plane index; Acquire an indoor point cloud of a building, perform supervoxel segmentation on the indoor point cloud of the building, merge adjacent supervoxels using a region growing algorithm, and generate an indoor multi-scale point cluster set with plane feature preservation, and extract an indoor plane structure point cluster from the indoor multi-scale point cluster set according to the calculated indoor point cluster plane index; The method of obtaining an outdoor point cloud of a building, performing supervoxel segmentation on the outdoor point cloud of the building, merging adjacent supervoxels using a region growing algorithm, and generating an outdoor multi-scale point cluster set with plane feature preservation, and extracting an outdoor plane structure point cluster from the outdoor multi-scale point cluster set according to the calculated outdoor point cluster plane index, specifically includes: The outdoor point cloud of the building is subjected to supervoxel segmentation using a heuristic supervoxel segmentation algorithm based on local information. The energy function used by the heuristic supervoxel segmentation algorithm is composed of a geometric difference term and a point cluster quantity term: Among them, E(Z) represents the entire energy function, z ij To determine the point p j Does it belong to the point p i In the supervoxel represented by, λ is the regularization coefficient, which is used to weigh the weight of the geometric difference term and the number of point clusters in the energy function, C(Z) represents the number of supervoxels currently generated, K represents the expected number of supervoxels obtained according to the preset resolution, and D(p i ,p j ) is a similarity measurement parameter used to measure the similarity between two points, and N represents the number of point clusters; Similarity measurement parameter D(p i ,p j ) is defined as: Among them, D s is the Euclidean distance between two points, D n is the normal vector n of the two points i and n j dot product of, R is the predefined supervoxel resolution, μ and ∈ are D s and D n The weight of Solving the optimal solution of the energy function to achieve supervoxel segmentation of the outdoor point cloud of the building; Supervoxels are selected as seed supervoxels in order of growth; Supervoxels are grown according to the neighborhood relationship between them, and only supervoxels with planar similarity are allowed to be segmented into the same point cluster; If no new supervoxel is segmented into the point cluster where the seed supervoxel is located, the next seed supervoxel is selected for growth until all supervoxels are grown; After all the planar point clusters are segmented, the unmerged supervoxels are merged with the planar point cluster set to generate an outdoor multi-scale point cluster set with plane feature preservation. If a multi-scale point cluster is C, the multi-scale point cluster C is fitted into a plane equation based on the least squares method: Π(x, y, z): Ax+By+C'z+D=0; (4) Among them, Π represents the plane structure, A, B, C' are the three coefficients of the plane equation, and D is a constant; The residual r is obtained by calculating the average distance from the midpoint of the multi-scale point cluster C to the fitting plane: Where n represents the number of multi-scale point clusters, and i represents any one of the multi-scale point clusters; If p i (x i ,y i , z i ) point in the cluster, obtain the eigenvalue to calculate the plane feature of the cluster, perform eigendecomposition on the covariance matrix Q of the multi-scale cluster C to obtain the eigenvalues ​​λ1, λ2, λ3 of the cluster, where λ1>λ2>λ3, and obtain the plane feature p of the cluster Π : Define the point cluster size c as the number of points contained in the point cluster. If the normalized point cluster size is: Among them, c max is the point cluster size of the largest point cluster in the outdoor multi-scale point cluster set; Calculate the point cluster plane index P Π : Set the threshold t according to the point cluster plane index P Π And threshold t extracts outdoor plane structure point clusters from the outdoor multi-scale point cluster set. If the point cluster plane index P Π If it is greater than the threshold t, it belongs to the outdoor plane structure point cluster.

2. The method for instantiating and extracting building components based on indoor and outdoor point clouds according to claim 1, characterized in that: The step of extracting the structural line segments of the building based on the outdoor structural plane point clusters and extracting the building floor heights and room corner points based on the indoor structural plane point clusters specifically includes: The edge straight line extraction technology is used to extract the edge straight lines of the outdoor plane structure point cluster to obtain multiple three-dimensional line segments; Arrange multiple three-dimensional line segments in descending order according to their lengths, cluster the multiple three-dimensional line segments according to their relative directions, and take out the three-dimensional line segment with the longest length as the starting reference line segment L of the first cluster lc1; Calculate the line segments to be clustered l in sequence according to the order of line segment arrangement i and reference line segment l r The angle between like It is believed that l i ∈lc1, if Then create a new cluster lc i , think l i is the clustering lc i The reference line segment; Repeat the clustering operation until all three-dimensional line segments are clustered. For the three-dimensional line segments that are not clustered, they are determined to be noise segments and removed; After completing the line segment clustering, calculate the sum of the line segment lengths in each cluster, and set the line segment clusters with the first two digits of the sum of the line segment lengths as the potential structural line segment set of the outdoor plane structure point cluster; According to the length of the line segment, the line segments without component information are eliminated, and the structural index St of the line segment cluster is used. Π Dynamically adjust the rejection threshold th; Among them, the structural index St Π The calculation is: calculate the angle between the line segments in the line segment set two by two like || means or, then the line segment l is considered i and l i+1 It has structurality, and the ratio of the sum of the lengths of the structural line segments to the sum of the lengths of all line segments in the line segment set is calculated as the structural index St of the outdoor plane structure point cluster. Π ; If the length of the longest line segment in the outdoor plane structure point cluster is L', dynamically adjust the elimination threshold th: If the length of the line segment in the outdoor plane structure point cluster is less than th, it will be eliminated; Calculate the remaining line segments l and the perpendicular line l in the structure line segment cluster ⊥ Angle of direction (0,0,1) like Then the line segment l is considered as a vertical structural line segment and retained, and the structural line segment of the building is obtained; According to the direction of the normal vector of the structural plane point cluster, the horizontal structural point cluster and the vertical structural point cluster are distinguished. According to the elevation of the horizontal point cluster, the ground bottom plate and the top plate point cluster are extracted. According to the average distance between the floor point cluster and the top plate point cluster, the building floor height f is calculated. The vertical point cluster is projected onto the plane where the ground point cluster is located to generate a binary mapping diagram of the indoor space. The morphological segmentation method and the regular grid division method are combined on the binary map to extract the two-dimensional room corner points; Map the two-dimensional room corner points back to three-dimensional space, and optimize the inaccurate extraction of room corner points caused by the shared wall problem. The optimization process is as follows: Take a room corner point p, search for neighboring points of corner point p with the thickness of the component t' as the neighborhood search radius, and construct a neighborhood point set N{p,p2,p3,…,p n }, calculate the center point p of the point set c As the final room corner point.

3. The method for instantiating and extracting building components based on indoor and outdoor point clouds according to claim 2, characterized in that: The spatial association relationships include: adjacent relationships, nested relationships, supporting relationships, connected relationships and suspended relationships; the geometric association relationships are geometric abstractions of spatial association relationships, including: parallel relationships, vertical relationships, aligned relationships, tangent relationships and separated relationships; the prior knowledge includes: first prior knowledge, second prior knowledge, third prior knowledge and fourth prior knowledge. The first prior knowledge is that there are column components at the intersection of walls of different structures, the second prior knowledge is that the columns and walls are perpendicular to the ground and the beams are horizontal to the ground, the third prior knowledge is that the top plate and the bottom plate are connected to the beams and suspended on the columns, and the fourth prior knowledge is that the walls, beams and columns are interconnected and located in the same plane structure.

4. The method for instantiating and extracting building components based on indoor and outdoor point clouds according to claim 3, characterized in that: The method of identifying the positions of components of different categories based on the outdoor structural plane point cluster and the indoor structural plane point cluster, taking the structural line segments, the building floor heights and the room corner points as position references, and combining the prior knowledge specifically includes: The position of the column component is identified, and according to the first prior knowledge and the second prior knowledge, it is concluded that the column component exists at the structural line segment extracted based on the plane structure of the outdoor point cloud of the building; According to the first prior knowledge, it is inferred that there are column components at the corners of the indoor rooms of the building. For each independent indoor space, a set of corner points P'{p'1,p'2…,p' n’ }, the points in the corner point set P' are sorted in clockwise order, and n' represents the number of corner points; If p' i is a point in P', according to p' i Construct two straight lines from the left and right corners of Find the straight line The intersection points with the plane structure Π of the building facade create an intersection point set Q{q1,q2…,q n” }, n" represents the number of intersection points; calculate p' i With the intersection point q i If the distance d is greater than the thickness t' of the component, then q i Remove from the intersection set Q, the remaining intersection points Q{q1,q2…,q m } represents the corner point p' i For the position of the column component on the plane structure Π, m represents the number of remaining intersections. According to the direction and length of the vertical structural line on the plane, a column identification line segment l is created with the intersection as the starting point. C ; After the position of the column component is identified, the beam component is identified according to the third prior knowledge and the fourth prior knowledge; If the beam identification line segment l B It is the position mark of the beam component on the plane structure. If there is an original column identification line segment set on the plane structure Π of the building The line segments in the set are arranged in a clockwise direction on the plane; Take the column identification line and in, They are the starting points of the two marking lines close to the ground. is the end point of the other end of the marking line; The distance between the two starting points along the vertical direction of the marking line is derived to obtain beam B. i The two end nodes and The beam identification line Node-based and Column identification line and Split and generate new column identification lines as well as And update the original identification line, after updating, it becomes and Until and The length of is less than the building floor height f, and every column identification line segment is included in the calculation; Use a horizontal plane Π consisting of four nodes W (n1, n2, n3, n4) identifies the location of the wall component. After determining the location of the column component and the beam component on the plane structure of the building, according to the fourth prior knowledge, it is known that the node of the wall component coincides with the intersection of the beam and column components. The node of the wall component is the first and last endpoints of two adjacent segmented column identification lines.

5. The method for instantiating and extracting building components based on indoor and outdoor point clouds according to claim 4, characterized in that: After obtaining the location of each category of components, generating a beam component bounding box, a column component bounding box and a wall component bounding box based on the component position identifier, and performing component point cloud segmentation on the building point cloud, specifically includes: The normal vector n of the plane structure Π along which the endpoint of the identified line segment lies Π Positive and negative direction derivatives The distance to get the vertex v′(v1,v2,v3,v4) of the component bounding box; Vertices n1, n2, n3, and n4 form a horizontal plane Π'. Let the normal vector of the horizontal plane Π' be n Π′ ; Identify the endpoints of the line segment along the normal vector n Π′ Positive and negative direction derivatives The distance of the remaining vertices v″(v5,v6,v7,v8) is obtained; Merge the vertex v′(v1,v2,v3,v4) of the component bounding box and the remaining vertex v″(v5,v6,v7,v8) to obtain the vertex V(v1,v2,v3,v4,v5,v6,v7,v8) of the component bounding box; Connect the vertices in the component bounding box vertices V (v1, v2, v3, v4, v5, v6, v7, v8) according to a preset connection method to form a beam component bounding box and a column component bounding box; If the wall component identification surface Π W The normal vector of (n1,n2,n3,n4) is Derive each node of the identified face along the positive and negative directions of the normal vector Get the vertex set of the component bounding box by Connect the vertices on the same horizontal plane and then connect the corresponding positive and negative direction nodes in sequence to generate the wall component bounding box; Compare the point p" in the point cloud with the center point p' of the component bounding box c To the different planes of the component bounding box i’ The difference in distance d i’ , the component bounding box includes a beam component bounding box, a column component bounding box and a wall component bounding box, and the building point cloud is segmented into component point clouds: d i’ =dis(plane i’ ,p’ c )-dis(plane i’ ,p”)s.t.{i’∈Z|1≤i’≤6};(10) Among them, dis(plane i’ ,p”) represents the distance from point p” in the point cloud to the i’th surface of the component, dis(plane i’ ,p' c ) represents the center point p' of the component bounding box c The distance to the i'th face of the component, Z represents the number of faces of the constructed bounding box; If all d in {i'∈Z|1≤i'≤6} i’ If both are less than 0, the current point is considered to belong to the component point cloud.

6. A building component instantiation extraction system based on indoor and outdoor point clouds, characterized in that: The building component instantiation extraction system based on indoor and outdoor point clouds includes: A multi-scale point cluster segmentation module is used to obtain indoor and outdoor point clouds of buildings, perform multi-scale point cluster segmentation on the indoor and outdoor point clouds of buildings, segment and separate different geometric structures in the building, calculate the point cluster plane index, and extract indoor structure plane point clusters and outdoor structure plane point clusters; A geometric element extraction module, used to extract the structural line segments of the building based on the outdoor structural plane point clusters, and to extract the building floor heights and room corner points based on the indoor structural plane point clusters; The correlation analysis module is used to analyze the spatial correlation and geometric correlation between components, and obtain the prior knowledge for identifying beams, columns and wall components in the building structure based on the spatial correlation and geometric correlation. A component position identification module, which is used to identify the positions of components of different categories based on the outdoor structural plane point cluster and the indoor structural plane point cluster, the structural line segments, the building floor heights and the room corner points as position references, and in combination with the prior knowledge; The component point cloud segmentation module is used to obtain the location of each category of components, generate beam component bounding boxes, column component bounding boxes and wall component bounding boxes based on the component position identification symbol, and perform component point cloud segmentation on the building point cloud; The method of obtaining indoor and outdoor point clouds of a building, performing multi-scale point cluster segmentation on the indoor and outdoor point clouds of the building, segmenting and separating different geometric structures in the building, calculating the point cluster plane index, and extracting indoor structure plane point clusters and outdoor structure plane point clusters specifically includes: Acquire an outdoor point cloud of a building, perform supervoxel segmentation on the outdoor point cloud of the building, merge adjacent supervoxels using a region growing algorithm, and generate an outdoor multi-scale point cluster set with plane feature preservation, and extract an outdoor plane structure point cluster from the outdoor multi-scale point cluster set according to the calculated outdoor point cluster plane index; Acquire an indoor point cloud of a building, perform supervoxel segmentation on the indoor point cloud of the building, merge adjacent supervoxels using a region growing algorithm, and generate an indoor multi-scale point cluster set with plane feature preservation, and extract an indoor plane structure point cluster from the indoor multi-scale point cluster set according to the calculated indoor point cluster plane index; The method of obtaining an outdoor point cloud of a building, performing supervoxel segmentation on the outdoor point cloud of the building, merging adjacent supervoxels using a region growing algorithm, and generating an outdoor multi-scale point cluster set with plane feature preservation, and extracting an outdoor plane structure point cluster from the outdoor multi-scale point cluster set according to the calculated outdoor point cluster plane index, specifically includes: The outdoor point cloud of the building is subjected to supervoxel segmentation using a heuristic supervoxel segmentation algorithm based on local information. The energy function used by the heuristic supervoxel segmentation algorithm is composed of a geometric difference term and a point cluster quantity term: Among them, E(Z) represents the entire energy function, z ij To determine the point p j Does it belong to the point p i In the supervoxel represented by, λ is the regularization coefficient, which is used to weigh the weight of the geometric difference term and the number of point clusters in the energy function, C(Z) represents the number of supervoxels currently generated, K represents the expected number of supervoxels obtained according to the preset resolution, and D(p i ,p j ) is a similarity measurement parameter used to measure the similarity between two points, and N represents the number of point clusters; Similarity measurement parameter D(p i ,p j ) is defined as: Among them, D s is the Euclidean distance between two points, D n is the normal vector n of the two points i and n j dot product of, R is the predefined supervoxel resolution, μ and ∈ are D s and D n The weight of Solving the optimal solution of the energy function to achieve supervoxel segmentation of the outdoor point cloud of the building; Supervoxels are selected as seed supervoxels in order of growth; Supervoxels are grown according to the neighborhood relationship between them, and only supervoxels with planar similarity are allowed to be segmented into the same point cluster; If no new supervoxel is segmented into the point cluster where the seed supervoxel is located, the next seed supervoxel is selected for growth until all supervoxels are grown; After all the planar point clusters are segmented, the unmerged supervoxels are merged with the planar point cluster set to generate an outdoor multi-scale point cluster set with plane feature preservation. If a multi-scale point cluster is C, the multi-scale point cluster C is fitted into a plane equation based on the least squares method: Π(x, y, z): Ax+By+C'z+D=0; (4) Among them, Π represents the plane structure, A, B, C' are the three coefficients of the plane equation, and D is a constant; The residual r is obtained by calculating the average distance from the midpoint of the multi-scale point cluster C to the fitting plane: Where n represents the number of multi-scale point clusters, and i represents any one of the multi-scale point clusters; If p i (x i ,y i , z i ) point in the cluster, obtain the eigenvalue to calculate the plane feature of the cluster, perform eigendecomposition on the covariance matrix Q of the multi-scale cluster C to obtain the eigenvalues ​​λ1, λ2, λ3 of the cluster, where λ1>λ2>λ3, and obtain the plane feature p of the cluster Π : Define the point cluster size c as the number of points contained in the point cluster. If the normalized point cluster size is: Among them, c max is the point cluster size of the largest point cluster in the outdoor multi-scale point cluster set; Calculate the point cluster plane index P Π : Set the threshold t according to the point cluster plane index P Π And threshold t extracts outdoor plane structure point clusters from the outdoor multi-scale point cluster set. If the point cluster plane index P Π If it is greater than the threshold t, it belongs to the outdoor plane structure point cluster.

7. A terminal, characterized in that: The terminal includes: a memory, a processor, and a building component instantiation extraction program based on indoor and outdoor point clouds stored in the memory and executable on the processor. When the building component instantiation extraction program based on indoor and outdoor point clouds is executed by the processor, the steps of the building component instantiation extraction method based on indoor and outdoor point clouds as described in any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a building component instantiation extraction program based on indoor and outdoor point clouds. When the building component instantiation extraction program based on indoor and outdoor point clouds is executed by a processor, the steps of the building component instantiation extraction method based on indoor and outdoor point clouds as described in any one of claims 1 to 5 are implemented.

Images