Building information model generation method, system and equipment based on unmanned aerial vehicle point cloud data

Abstract

The invention relates to a building information model generation method, system and device based on unmanned aerial vehicle point cloud data, and the method comprises the steps: building a building point cloud model based on a multi-view building image, and carrying out the neighborhood distance adaptive denoising; solving a curvature value of the denoised building point cloud model, and generating an optimal simplified point cloud by implementing curvature difference driven progressive point cloud deletion in a local neighborhood; dynamically constructing a triangular chain table structure according to the spatial distribution characteristics of the optimal simplified point cloud, and generating a building triangular mesh model by iteratively inserting the point cloud and reconstructing a topological connection relationship; and identifying and extracting a plane structure boundary and an embedded structure boundary of the target building by analyzing vertex topological information of the triangular mesh model, and mapping boundary information to generate a structured building information model. According to the method and the device, the time required for creating the building information model through the point cloud data is shortened, the modeling error-tolerant rate is improved, and the building information model of the target building is quickly and accurately created.

Description

Technical Field

[0001] The present invention relates to the technical field of building information model creation, and in particular to a method, system and device for generating a building information model based on drone point cloud data. Background Art

[0002] In recent years, point cloud data has been increasingly used in the construction industry. Its practical applications encompass a wide range of areas, including 3D scene modeling, geometric quality monitoring, project progress and safety management, and structural mechanics analysis. There are several main methods for acquiring 3D laser point clouds of buildings, depending on the operating mode of the scanning equipment: airborne laser scanning, gantry-mounted 3D laser scanning, and mobile 3D laser scanning. Drones can carry laser scanning equipment for scanning operations, and can also generate models and point clouds from massive amounts of photos and aerial triangulation results through oblique photography, offering the advantages of efficiency, flexibility, and cost-effectiveness.

[0003] Building Information Modeling (BIM) has gained widespread recognition within the Architecture, Engineering, and Construction (AEC) industry due to its applicability to the entire building lifecycle. Construction professionals require precise 3D models for projects such as building renovation, building performance analysis, and heritage preservation. However, detailed building information for some existing buildings is difficult to obtain due to lost or missing architectural drawings. Compared to traditional methods, creating BIM models of existing buildings using point cloud data can rapidly obtain building data, reduce human error, and save significant manpower and time. This process is often referred to as scan-to-BIM.

[0004] Currently, many software tools exist to assist users in the scan-to-BIM process. These tools help users import point cloud data into modeling tools, providing the necessary building structure and dimensional information for the creation of the target building's BIM. However, the modeling process is still manual, requiring users to master the use of these software tools.

[0005] Although the existing scan-to-BIM method can already provide assistance for BIM creation, there is still a lack of an automated point cloud-based BIM model generation method. Summary of the Invention

[0006] (1) Technical issues to be resolved

[0007] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a method, system and device for generating a building information model based on drone point cloud data, which solves the technical problem of low automation level of the existing point cloud-based BIM model generation method.

[0008] (2) Technical solution

[0009] In order to achieve the above objectives, the main technical solutions adopted by the present invention include:

[0010] In a first aspect, an embodiment of the present invention provides a method for generating a building information model based on drone point cloud data, comprising:

[0011] A building point cloud model is constructed based on multi-view building images acquired through drone oblique photography, and neighborhood distance adaptive denoising is applied;

[0012] Obtain the curvature value of the denoised building point cloud model and generate the optimal simplified point cloud by implementing curvature difference-driven progressive point cloud reduction in the local neighborhood.

[0013] Dynamically construct a triangle linked list structure based on the spatial distribution characteristics of the optimal simplified point cloud, and generate a geometrically continuous building triangular mesh model by iteratively inserting the point cloud and reconstructing the topological connection relationship;

[0014] By analyzing the vertex topology information of the triangular mesh model, the plane structure boundary of the target building is identified and extracted. At the same time, the region growing algorithm is combined with the normal vector similarity criterion to extract the embedded structure boundary. The plane structure boundary and the embedded structure boundary are mapped to generate a structured building information model.

[0015] Optionally, constructing a building point cloud model based on multi-view building images acquired through drone oblique photography and applying neighborhood distance adaptive denoising includes:

[0016] Obtain image data of target buildings from multiple viewpoints through drone oblique photography;

[0017] Based on 3D reconstruction software, perform point cloud computing on the image data to generate a point cloud model of the target building;

[0018] Apply neighborhood distance adaptive denoising processing to the building point cloud model to obtain the denoised building point cloud model.

[0019] Optionally, the curvature value of the denoised building point cloud model is obtained, and the optimal simplified point cloud is generated by performing curvature difference-driven progressive point cloud reduction in a local neighborhood, including:

[0020] Traverse all scattered points in the denoised building point cloud model and calculate the curvature value of each scattered point based on the local neighborhood geometric features;

[0021] Compare the curvature values ​​of each scattered point with the adjacent scattered points in the corresponding neighborhood, and select the original shape features of the target building based on the comparison results;

[0022] When the curvature value of the current scattered point is higher than the preset neighborhood curvature threshold, the current scattered point is determined to be the original shape feature point;

[0023] Traverse all original shape feature points, generate the optimal simplified point cloud and reconstruct the building point cloud model that retains the original geometric features.

[0024] Optionally, dynamically constructing a triangle linked list structure based on the spatial distribution characteristics of the optimal simplified point cloud, iteratively inserting the point cloud and reconstructing the topological connection relationship to generate a geometrically continuous building triangular mesh model includes:

[0025] Construct the initial triangle linked list structure based on the super triangle formed by all scattered points in the optimal simplified point cloud;

[0026] Traverse all scattered points and insert them into the initial triangle linked list structure point by point, and dynamically locate all the influencing triangles including the current insertion point based on the position information of the inserted scattered points;

[0027] By removing the common edges that affect the formation of triangles and generating cavity polygons, the inserted scattered points are topologically connected with the vertices of the cavity polygons to generate the optimal super triangle of the current scattered points;

[0028] The initial triangle linked list structure is iteratively updated according to the optimal super triangle to obtain a building triangular mesh model with geometric continuity characteristics.

[0029] Optionally, by parsing the vertex topology information of the triangular mesh model, identifying and extracting the plane structure boundary of the target building, and simultaneously combining the region growing algorithm with the normal vector similarity criterion to extract the embedded structure boundary, the plane structure boundary and the embedded structure boundary are mapped to generate a structured building information model, including:

[0030] Based on the plane segmentation algorithm, the plane structure of the triangular mesh model is extracted by combining the vertex coordinates and corresponding triangle information of the triangular mesh model. The plane structure boundary is obtained by performing boundary analysis on the normal vector information and geometric features of the plane structure.

[0031] Based on the region growing algorithm, the seed points selected in the plane structure are iteratively grown using the normal vector similarity criterion, and the embedded structure boundary is extracted based on the position and size information of the embedded structure of the obtained triangular mesh model;

[0032] The target building is decomposed into building information objects according to the plane structure boundary and embedded structure boundary, and a structured building information model is constructed by combining the geometric parameter information of the building information objects.

[0033] Optionally, based on a region growing algorithm, the seed points selected in the planar structure are iteratively grown using a normal vector similarity criterion, and the embedded structure boundary is extracted based on the position and size information of the embedded structure of the obtained triangular mesh model, including:

[0034] Based on the plane segmentation algorithm, the plane structure of the triangular mesh model is extracted by combining the vertex coordinates and the corresponding triangle information of the triangular mesh model;

[0035] Classify the plane structure according to its normal vector information and geometric features, and determine the building type corresponding to the plane structure;

[0036] The boundary information of each type of plane structure is obtained by performing boundary analysis on the classified plane structures and extracting the boundary line of each plane structure.

[0037] Optionally, based on a region growing algorithm, combined with seed points selected in the plane structure, the region is iteratively grown using a normal vector angle threshold, and the embedded structure boundary is obtained based on the position and size information of the embedded structure of the target building, including:

[0038] In the case where the plane structure is a wall structure, a seed point is selected as the starting point for growth;

[0039] Perform regional growth from the starting point in the direction of the seed point's normal vector, and traverse to calculate the angle between the normal vectors of the adjacent points and the normal vector of the current region;

[0040] When the angle between the normal vector of the adjacent point and the normal vector of the current region meets the set threshold condition, the adjacent point is included in the current region and updated as a growth point;

[0041] The region growing is iteratively performed until the set stop growing condition is met, and the position and size information of the embedded structure are obtained according to the three-dimensional point cloud coordinates of the obtained growing region.

[0042] Optionally, by parsing the vertex topology information of the triangular mesh model, identifying and extracting the plane structure boundary of the target building, and simultaneously extracting the embedded structure boundary by combining the region growing algorithm with the normal vector similarity criterion, and mapping the plane structure boundary with the embedded structure boundary to generate a structured building information model, the method further includes:

[0043] Compare the geometric parameters of each model component in the structured building information model with the actual geometric parameters of the building components to obtain the error values ​​of the model components in spatial position, geometric size and direction angle;

[0044] When the error value exceeds a preset threshold, the structured building information model is locally corrected based on the geometric parameter difference corresponding to the error value;

[0045] The error value of the direction angle includes: the angle deviation between the normal vector of the model component and the normal vector of the point cloud reference plane.

[0046] In a second aspect, an embodiment of the present invention provides a building information model generation system based on drone point cloud data, comprising:

[0047] A point cloud model generation module is used to construct a building point cloud model based on multi-view building images acquired through drone oblique photography and apply neighborhood distance adaptive denoising;

[0048] The point cloud model simplification module is used to obtain the curvature value of the denoised building point cloud model and generate the optimal simplified point cloud by implementing progressive point cloud reduction driven by curvature differences in the local neighborhood;

[0049] The triangular mesh model construction module is used to dynamically construct a triangle linked list structure based on the spatial distribution characteristics of the optimal simplified point cloud. By iteratively inserting the point cloud and reconstructing the topological connection relationship, a geometrically continuous building triangular mesh model is generated.

[0050] The building information model construction module is used to identify and extract the plane structure boundaries of the target building by parsing the vertex topology information of the triangular mesh model. At the same time, it combines the region growing algorithm with the normal vector similarity criterion to extract the embedded structure boundaries, and maps the plane structure boundaries with the embedded structure boundaries to generate a structured building information model.

[0051] In a third aspect, an embodiment of the present invention provides a building information model generation device, comprising:

[0052] A drone shooting device is used to perform multi-viewpoint oblique photography and transmit a real-time image stream of the target building surface;

[0053] Wireless communication module, used to provide a real-time point cloud transmission link between the drone camera and the processor;

[0054] The processor is connected to the drone shooting device through the wireless communication module and is used to execute the above-mentioned building information model generation method based on drone point cloud data.

[0055] (3) Beneficial effects

[0056] The beneficial effects of the present invention are as follows: the method, system, and device for generating a building information model based on drone point cloud data utilize drone oblique photography technology, point cloud extraction, and processing, and other artificial intelligence algorithms to automatically generate a building point cloud, reconstruct the building point cloud into a triangular mesh model, and then generate a building information model from the triangular mesh model, thereby achieving automated generation of a building information model from a point cloud. Compared to existing technologies, the present invention significantly reduces the time required to manually create a building information model from point cloud data, does not require the user to possess excessive professional knowledge, reduces the error-prone modeling process, and can quickly and accurately create a building information model of the target building. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1A schematic diagram of a flow chart of a method for generating a building information model based on drone point cloud data provided by one embodiment of the present invention;

[0058] Figure 2 A simplified flowchart of a method for generating a building information model based on drone point cloud data provided by one embodiment of the present invention;

[0059] Figure 3 、 Figure 4 Multi-viewpoint image data obtained through oblique photography by a drone provided in one embodiment of the present invention;

[0060] Figure 5 A building point cloud model diagram provided by an embodiment of the present invention;

[0061] Figure 6 A point cloud image with outlier noise provided by an embodiment of the present invention;

[0062] Figure 7 A denoised point cloud image provided by an embodiment of the present invention;

[0063] Figure 8 A schematic diagram of the Delaunay triangulation algorithm provided in one embodiment of the present invention;

[0064] Figure 9 A triangular mesh model diagram of a building provided by an embodiment of the present invention;

[0065] Figure 10 A structured building information model diagram is provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0066] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation methods in conjunction with the accompanying drawings.

[0067] refer to Figure 1 and Figure 2As shown, an embodiment of the present invention proposes a method for generating a building information model based on drone point cloud data, the method comprising: first, constructing a building point cloud model based on multi-viewpoint building images obtained through drone oblique photography, and applying neighborhood distance adaptive denoising; then, obtaining the curvature value of the denoised building point cloud model, and generating an optimal simplified point cloud by implementing curvature difference-driven progressive point cloud pruning in a local neighborhood; secondly, dynamically constructing a triangle linked list structure according to the spatial distribution characteristics of the optimal simplified point cloud, and generating a geometrically continuous building triangular mesh model by iteratively inserting point clouds and reconstructing topological connection relationships; finally, by parsing the vertex topology information of the triangular mesh model, identifying and extracting the plane structure boundary of the target building, and at the same time combining the region growing algorithm with the normal vector similarity criterion to extract the embedded structure boundary, and mapping the plane structure boundary with the embedded structure boundary to generate a structured building information model.

[0068] This embodiment utilizes drone oblique photography technology, point cloud extraction, and processing, along with other artificial intelligence algorithms, to automatically generate building point clouds. This point cloud is then reconstructed into a triangular mesh model, which is then used to generate a Building Information Model (BIM), achieving automated conversion from point clouds to BIMs. Compared to existing technologies, this embodiment significantly reduces the time required to manually create a BIM from point cloud data, eliminates the need for extensive user expertise, and reduces the risk of errors during the modeling process, enabling the rapid and accurate creation of a BIM for the target building.

[0069] To better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. Rather, these embodiments are provided to enable a clearer and more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0070] Specifically, refer to Figure 1 As shown, this embodiment proposes a method for generating a building information model based on drone point cloud data, the method comprising:

[0071] S100: Build a building point cloud model based on multi-view building images obtained through drone oblique photography, and apply neighborhood distance adaptive denoising.

[0072] In this embodiment, step S100 may include the following sub-steps S110 to S130:

[0073] S110 , acquiring image data of a target building from multiple viewpoints through oblique photography by a drone.

[0074] Preferably, a five-lens drone oblique photography system (1 vertical + 4 oblique lenses) or a flight path planning method is used to plan spiral or cross routes around the target building. By setting a heading / side overlap of 70%-80%, it is ensured that each building part is covered by at least 5 perspectives at different heights and angles. Compared with traditional orthophotos, oblique photography can simultaneously obtain the target building facade texture (such as window structure), roof details (cornice shape) and ground attachments (awnings, steps). The parallax information formed by multi-viewpoint images provides key geometric clues for subsequent three-dimensional reconstruction. For example, multi-viewpoint image data such as Figure 3 and 4 shown.

[0075] S120. Perform point cloud computing on the image data based on 3D reconstruction software to generate a building point cloud model of the target building.

[0076] For example, using 3D reconstruction software (such as Agisoft Metescan, RealityCapture, etc.) to analyze Figure 3 and Figure 4 The disparity information of the multi-view images shown in the figure is used to solve the spatial coordinates using the collinear equation in photogrammetry to generate the following Figure 5 The architectural point cloud model of the target building is shown.

[0077] S130 , applying neighborhood distance adaptive denoising processing to the building point cloud model to obtain a denoised building point cloud model.

[0078] Preferably, the noise removal and smoothing of the building point cloud model is performed by Gaussian filtering algorithm. For example, first, the point cloud data of the building point cloud model to be processed is read, such as Figure 6 As shown in the figure, the building point cloud model contains outlier noise. Next, set the Gaussian filter parameters, including the kernel size (i.e., neighborhood range) and the standard deviation of the Gaussian function. Then, for each point in the point cloud, calculate the new coordinate value G(x,y) based on the points in its neighborhood and the Gaussian weight function. Finally, save the filtered point cloud data and obtain Figure 7 The denoised building point cloud model is shown.

[0079]

[0080] In formula (1), σ is the standard deviation of the Gaussian function.

[0081] S200, obtaining the curvature value of the denoised building point cloud model, and generating an optimal simplified point cloud by implementing curvature difference-driven progressive point cloud reduction in a local neighborhood.

[0082] In this embodiment, a curvature adaptive simplification algorithm is used to simplify the building point cloud model. While maintaining the geometric characteristics of the target building, the sampling density is dynamically adjusted according to the curvature size to achieve streamlined optimization of the point cloud data, reduce the complexity and redundancy of the point cloud data, and thus shorten the generation time of the subsequent building information model.

[0083] In this embodiment, step S200 may include the following sub-steps S210 to S240:

[0084] S210: traverse all scattered points in the denoised building point cloud model, and obtain the curvature value of each scattered point based on the local neighborhood geometric features.

[0085] S220: Compare the curvature values ​​of each scattered point with those of adjacent scattered points in the corresponding neighborhood, and select the original shape features of the target building based on the comparison results.

[0086] S230: When the curvature value of the current scatter point is higher than a preset neighborhood curvature threshold, the current scatter point is determined to be an original shape feature point.

[0087] S240, traversing all original shape feature points, generating an optimal simplified point cloud and reconstructing a building point cloud model that retains the original geometric features.

[0088] Preferably, first, the curvature value of each scattered point is calculated. Then, based on the calculated curvature values, the point cloud data is filtered and simplified. By comparing the curvature values ​​of adjacent points, key points with higher curvature are retained while redundant points with lower curvature are deleted, thereby reducing the data volume while retaining the main geometric features. Finally, the point cloud data is reconstructed based on the simplified point set to obtain a more streamlined building point cloud model that still retains the original shape characteristics.

[0089] S300, dynamically constructing a triangle linked list structure according to the spatial distribution characteristics of the optimal simplified point cloud, and generating a geometrically continuous building triangular mesh model by iteratively inserting the point cloud and reconstructing the topological connection relationship.

[0090] In this embodiment, the Figure 8 The Delaunay triangulation algorithm shown reconstructs the building point cloud model into Figure 9 The building triangulated mesh model shown above is used to generate a Building Information Model (BIM), further reducing its generation time while ensuring BIM accuracy. An incremental insertion strategy reduces topological errors by 87% compared to traditional batch processing, improving topological robustness. Furthermore, the Delaunay triangulation algorithm offers multi-scale modeling capabilities, supporting multiple levels of expression from LOD1 (whole volumes) to LOD4 (detailed components). A single model can support multiple scales from 1:1000 to 1:50.

[0091] In this embodiment, step S300 may include the following sub-steps S310 to S340:

[0092] S310: Construct an initial triangle linked list structure based on the super triangle formed by all scattered points in the optimal simplified point cloud.

[0093] S320: traverse all scattered points and insert them into the initial triangle linked list structure point by point, and dynamically locate all influencing triangles including the current insertion point based on the position information of the inserted scattered points.

[0094] For example, insert the scattered points P in the point set into the initial triangle linked list structure in sequence, refer to Figure 8 Next, refer to Figure 8 In Figure (b), determine the circumscribed circle containing the inserted scattered point P, and find the affected triangles (triangle ABC and triangle ABD) whose circumscribed circle contains the inserted scattered point P in the initial triangle linked list structure.

[0095] S330: Remove the common edges that affect the formation of the triangle and generate a cavity polygon, topologically connect the inserted scattered point with each vertex of the cavity polygon, and generate the optimal super triangle of the current scattered point.

[0096] For example, remove the common edge AB that affects triangle ABC and triangle ABD, and generate a cavity polygon ABCD, such as Figure 8 As shown in Figure (c), the inserted scattered point P is then topologically connected with the vertices ABCD of the cavity polygon to generate the optimal super triangle (triangle APC, triangle CBP, triangle BDP, and triangle DAP) of the current scattered point.

[0097] S340. Iteratively update the initial triangle linked list structure based on the optimal super triangle to obtain a building triangle mesh model with geometric continuity features.

[0098] S400: Identify and extract the plane structure boundary of the target building by analyzing the vertex topology information of the triangular mesh model, and extract the embedded structure boundary by combining the region growing algorithm with the normal vector similarity criterion, and map the plane structure boundary with the embedded structure boundary to generate a structured building information model.

[0099] In this embodiment, step S400 may include the following sub-steps S410 to S430:

[0100] S410 , based on a plane segmentation algorithm, combined with vertex coordinates and corresponding triangle information of the triangular mesh model, extract the plane structure of the triangular mesh model, and perform boundary analysis on the normal vector information and geometric features of the plane structure to obtain the plane structure boundary.

[0101] In this embodiment, step S410 may include the following sub-steps S411 to S413:

[0102] S411. Based on a plane segmentation algorithm, the plane structure of the triangular mesh model is extracted by combining the vertex coordinates of the triangular mesh model and the corresponding triangle information.

[0103] S412. Classify the plane structure according to the normal vector information and geometric features of the plane structure, and determine the building type corresponding to the plane structure.

[0104] S413. Perform boundary analysis on the classified plane structures to obtain boundary information of each type of plane structure and extract the boundary line of each plane structure.

[0105] Preferably, the triangular mesh model is processed using a plane segmentation algorithm based on the vertex coordinates in the triangular mesh model and the triangle information connecting these vertices to identify the plane structures such as walls and floor slabs. The normal vector is calculated for each plane structure to determine the orientation and direction of the plane. According to the normal vector information and the geometric characteristics of the plane, the identified plane structures are classified, such as identifying horizontal planes as floor slabs and vertical planes as walls. Based on the boundary information of the identified floor slabs and walls, the boundary line of each plane is extracted. The plane segmentation algorithm improves the speed of extracting plane structures in the building information model, and also supports the automatic generation of building components that meet the IFC standard, thereby improving the efficiency of converting three-dimensional scans to BIM models.

[0106] S420: Based on a region growing algorithm, the seed points selected in the plane structure are iteratively grown using a normal vector similarity criterion, and the embedded structure boundary is extracted based on the position and size information of the embedded structure of the obtained triangular mesh model.

[0107] In this embodiment, step S420 may include the following sub-steps S421 to S424:

[0108] S421. When the plane structure is a wall structure, a seed point is selected as a growth starting point.

[0109] S422: Perform regional growth on the growth starting point in the direction of the normal vector of the seed point, and traverse and calculate the angle between the normal vectors of adjacent points and the normal vector of the current region.

[0110] S423: When the angle between the normal vector of the adjacent point and the normal vector of the current region meets a set threshold condition, the adjacent point is included in the current region and updated as a growth point.

[0111] S424 , iteratively executing region growing until a set stop growing condition is met, and obtaining position information and size information of the embedded structure according to the obtained three-dimensional point cloud coordinates of the growing region.

[0112] Preferably, since doors and windows are often embedded in building walls, a seed point is selected as the starting point in the planar structure of the wall. Next, the normal information of the region is checked as a similarity criterion. The neighboring points of the seed point are traversed and their similarity with the current region is calculated. If the neighboring point meets the similarity criteria, it is added to the current region and used as a new seed point. The above steps are repeated until no new points can be added to the region. Finally, the position and size information of the embedded structure are obtained based on the obtained 3D point cloud coordinates of the growth region.

[0113] S430: Decompose the target building into building information objects based on the plane structure boundary and the embedded structure boundary, and construct a structured building information model by combining the geometric parameter information of the building information objects.

[0114] Preferably, the target building is decomposed into different BIM objects, such as walls, floors, windows, etc., based on the plane structure boundary and the embedded structure boundary. Finally, the geometric parameter information of the walls, floors, windows, etc. is obtained and the structured building information model of the target building is generated in stages. Figure 10 shown.

[0115] In this embodiment, after step S400, the following steps S510 to S520 may be further included:

[0116] S510: Compare the geometric parameters of each model component in the structured building information model with the actual geometric parameters of the building component to obtain spatial coordinate errors in the spatial position, geometric dimensions, and orientation angles of the model component. The orientation angle error includes the angle deviation between the normal vector of the model component and the normal vector of the point cloud reference plane.

[0117] S520: When the error value exceeds a preset threshold, locally modify the structured building information model based on the geometric parameter difference corresponding to the error value.

[0118] To ensure the accuracy of the automatically generated structured building information model, a deviation analysis can be performed on the generated structured building information model. Specifically, by comparing the geometry in the structured building information model with the geometry of the actual building structure represented by the point cloud data, the shapes of elements such as the building's walls, roof, doors, and windows are checked to ensure they match the actual structure. This step is crucial in determining the positional, geometric, and directional errors of the model, thereby verifying the feasibility of the proposed method for automatically creating a structured building information model from point cloud data and ensuring the practical application of the generated BIM model.

[0119] In addition, this embodiment proposes a building information model generation system based on drone point cloud data, which includes:

[0120] The point cloud model generation module is used to construct a building point cloud model based on multi-view building images obtained through drone oblique photography and apply neighborhood distance adaptive denoising.

[0121] The point cloud model simplification module is used to obtain the curvature value of the denoised building point cloud model and generate the optimal simplified point cloud by implementing progressive point cloud reduction driven by curvature differences in the local neighborhood.

[0122] The triangular mesh model construction module is used to dynamically construct a triangle linked list structure based on the spatial distribution characteristics of the optimal simplified point cloud, and generate a geometrically continuous building triangular mesh model by iteratively inserting the point cloud and reconstructing the topological connection relationship.

[0123] The building information model construction module is used to identify and extract the plane structure boundaries of the target building by parsing the vertex topology information of the triangular mesh model. At the same time, it combines the region growing algorithm with the normal vector similarity criterion to extract the embedded structure boundaries, and maps the plane structure boundaries with the embedded structure boundaries to generate a structured building information model.

[0124] Furthermore, this embodiment provides a building information model generation device, including: a drone shooting device, a wireless communication module, and a wireless communication module processor.

[0125] The drone shooting device is used to perform multi-viewpoint oblique photography and transmit the target building surface image stream in real time; the wireless communication module is used to provide a real-time point cloud transmission link between the drone shooting device and the processor; the processor is connected to the drone shooting device through the wireless communication module and is used to execute the above-mentioned building information model generation method based on drone point cloud data.

[0126] In summary, this example proposes a method, system, and device for generating a building information model based on drone point cloud data. To address the problem of existing buildings lacking detailed architectural drawings and accurate geometric information, this embodiment further improves the current point cloud-based building information model creation method and provides a set of fully automatic generation technology routes based on point cloud data to obtain the combined information of the target building. First, the building point cloud is automatically generated using drone oblique photography technology, point cloud extraction and processing, and other artificial intelligence algorithms. The building point cloud is then reconstructed into a triangular mesh model. Finally, the triangular mesh model is used to generate a building information model, realizing the automated generation of the point cloud to building information model. This embodiment greatly reduces the time required to manually create a building information model from point cloud data and does not require the user to master excessive professional knowledge, making the modeling process less prone to errors and allowing the target building's building information model to be created quickly and accurately. At the same time, because the generated building information model has excellent accuracy, it can also be well applied to important projects such as building renovation, building performance analysis, and site protection.

[0127] Since the systems / devices described in the above embodiments of the present invention are systems / devices used to implement the methods of the above embodiments of the present invention, those skilled in the art will be able to understand the specific structures and variations of these systems / devices based on the methods described in the above embodiments of the present invention, and thus will not be described in detail here. All systems / devices used in the methods of the above embodiments of the present invention are within the scope of protection of the present invention.

[0128] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0129] The present invention is described with reference to flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions.

[0130] It should be noted that, in the description of the present invention, the word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The present invention can be implemented by means of hardware comprising several distinct components and by means of a suitably programmed computer. The use of the words first, second, third, etc., is merely for convenience and does not imply any order. These words should be understood as part of the component name.

[0131] In addition, it should be noted that, in the description of this specification, the description of the terms "one embodiment", "some embodiments", "embodiment", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0132] Although preferred embodiments of the present invention have been described, those skilled in the art will readily appreciate additional variations and modifications to these embodiments upon obtaining the basic inventive concepts.

[0133] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the invention.

Claims

1. A method for generating a building information model based on drone point cloud data, characterized in that: include: A building point cloud model is constructed based on multi-view building images acquired through drone oblique photography, and neighborhood distance adaptive denoising is applied; Obtain the curvature value of the denoised building point cloud model and generate the optimal simplified point cloud by implementing curvature difference-driven progressive point cloud reduction in the local neighborhood. Dynamically construct a triangle linked list structure based on the spatial distribution characteristics of the optimal simplified point cloud, and generate a geometrically continuous building triangular mesh model by iteratively inserting the point cloud and reconstructing the topological connection relationship; By analyzing the vertex topology information of the triangular mesh model, the plane structure boundary of the target building is identified and extracted. At the same time, the region growing algorithm is combined with the normal vector similarity criterion to extract the embedded structure boundary. The plane structure boundary and the embedded structure boundary are mapped to generate a structured building information model.

2. The method according to claim 1, wherein Building point cloud models are constructed based on multi-view building images acquired through drone oblique photography, and neighborhood distance adaptive denoising is applied, including: Obtain image data of target buildings from multiple viewpoints through drone oblique photography; Based on 3D reconstruction software, perform point cloud computing on the image data to generate a building point cloud model of the target building; Apply neighborhood distance adaptive denoising processing to the building point cloud model to obtain the denoised building point cloud model.

3. The method according to claim 1, wherein Obtain the curvature value of the denoised building point cloud model and generate the optimal simplified point cloud by implementing curvature difference-driven progressive point cloud reduction within the local neighborhood. Traverse all scattered points in the denoised building point cloud model and calculate the curvature value of each scattered point based on the local neighborhood geometric features; Compare the curvature values ​​of each scattered point with the adjacent scattered points in the corresponding neighborhood, and select the original shape features of the target building based on the comparison results; When the curvature value of the current scattered point is higher than the preset neighborhood curvature threshold, the current scattered point is determined to be the original shape feature point; Traverse all original shape feature points, generate the optimal simplified point cloud and reconstruct the building point cloud model that retains the original geometric features.

4. The method according to claim 1, wherein Dynamically construct a triangle linked list structure based on the spatial distribution characteristics of the optimal simplified point cloud. By iteratively inserting the point cloud and reconstructing the topological connection relationship, a geometrically continuous building triangular mesh model is generated, including: Construct the initial triangle linked list structure based on the super triangle formed by all scattered points in the optimal simplified point cloud; Traverse all scattered points and insert them into the initial triangle linked list structure point by point, and dynamically locate all the influencing triangles including the current insertion point based on the position information of the inserted scattered points; By removing the common edges that affect the formation of triangles and generating cavity polygons, the inserted scattered points are topologically connected with the vertices of the cavity polygons to generate the optimal super triangle of the current scattered points; The initial triangle linked list structure is iteratively updated according to the optimal super triangle to obtain a building triangular mesh model with geometric continuity characteristics.

5. The method according to claim 1, wherein By analyzing the vertex topology information of the triangular mesh model, the plane structure boundary of the target building is identified and extracted. At the same time, the region growing algorithm is combined with the normal vector similarity criterion to extract the embedded structure boundary. The plane structure boundary and the embedded structure boundary are mapped to generate a structured building information model, including: Based on the plane segmentation algorithm, the plane structure of the triangular mesh model is extracted by combining the vertex coordinates and corresponding triangle information of the triangular mesh model. The plane structure boundary is obtained by performing boundary analysis on the normal vector information and geometric features of the plane structure. Based on the region growing algorithm, the seed points selected in the plane structure are iteratively grown using the normal vector similarity criterion, and the embedded structure boundary is extracted based on the position and size information of the embedded structure of the obtained triangular mesh model; The target building is decomposed into building information objects according to the plane structure boundary and embedded structure boundary, and a structured building information model is constructed by combining the geometric parameter information of the building information objects.

6. The method according to claim 5, wherein Based on the region growing algorithm, the seed points selected in the plane structure are iteratively grown using the normal vector similarity criterion, and the embedded structure boundary is extracted based on the position and size information of the embedded structure of the obtained triangular mesh model, including: Based on the plane segmentation algorithm, the plane structure of the triangular mesh model is extracted by combining the vertex coordinates and the corresponding triangle information of the triangular mesh model; Classify the plane structure according to its normal vector information and geometric features, and determine the building type corresponding to the plane structure; The boundary information of each type of plane structure is obtained by performing boundary analysis on the classified plane structures and extracting the boundary line of each plane structure.

7. The method according to claim 5, wherein Based on the region growing algorithm, combined with the seed points selected in the plane structure, the region is iteratively grown through the normal vector angle threshold, and the embedded structure boundary is obtained based on the position and size information of the embedded structure of the target building. In the case where the plane structure is a wall structure, a seed point is selected as the starting point for growth; Perform regional growth from the starting point in the direction of the seed point's normal vector, and traverse to calculate the angle between the normal vectors of the adjacent points and the normal vector of the current region; When the angle between the normal vector of the adjacent point and the normal vector of the current region meets the set threshold condition, the adjacent point is included in the current region and updated as a growth point; The region growing is iteratively performed until the set stop growing condition is met, and the position and size information of the embedded structure are obtained according to the three-dimensional point cloud coordinates of the obtained growing region.

8. The method according to claim 1, wherein By analyzing the vertex topology information of the triangular mesh model, the plane structure boundary of the target building is identified and extracted. At the same time, the region growing algorithm is combined with the normal vector similarity criterion to extract the embedded structure boundary. After mapping the plane structure boundary with the embedded structure boundary to generate a structured building information model, the following is also included: Compare the geometric parameters of each model component in the structured building information model with the actual geometric parameters of the building components to obtain the error values ​​of the model components in spatial position, geometric size and direction angle; When the error value exceeds a preset threshold, the structured building information model is locally corrected based on the geometric parameter difference corresponding to the error value; The error value of the direction angle includes: the angle deviation between the normal vector of the model component and the normal vector of the point cloud reference plane.

9. A building information model generation system based on drone point cloud data, characterized in that: include: A point cloud model generation module is used to construct a building point cloud model based on multi-view building images acquired through drone oblique photography and apply neighborhood distance adaptive denoising; The point cloud model simplification module is used to obtain the curvature value of the denoised building point cloud model and generate the optimal simplified point cloud by implementing progressive point cloud reduction driven by curvature differences in the local neighborhood; The triangular mesh model construction module is used to dynamically construct a triangle linked list structure based on the spatial distribution characteristics of the optimal simplified point cloud. By iteratively inserting the point cloud and reconstructing the topological connection relationship, a geometrically continuous building triangular mesh model is generated. The building information model construction module is used to identify and extract the plane structure boundaries of the target building by parsing the vertex topology information of the triangular mesh model. At the same time, it combines the region growing algorithm with the normal vector similarity criterion to extract the embedded structure boundaries, and maps the plane structure boundaries with the embedded structure boundaries to generate a structured building information model.

10. A building information model generation device, characterized in that: include: A drone shooting device is used to perform multi-viewpoint oblique photography and transmit a real-time image stream of the target building surface; Wireless communication module, used to provide a real-time point cloud transmission link between the drone camera and the processor; The processor is connected to the drone shooting device through a wireless communication module, and is used to execute the building information model generation method based on drone point cloud data described in any one of claims 1 to 8.

Images