Method and system for automatic generation of building plan based on three-dimensional point cloud

Abstract

The application discloses a kind of based on three-dimensional point cloud's building plan automatically generated method and system, the building plan automatically generated method includes: obtaining the three-dimensional point cloud data of a residence;The three-dimensional point cloud data is preprocessed to obtain residence model;Residence model is projected and sliced;Region growth segmentation is carried out to projection slice image;Topology structure between the straight line fitted in projection slice is established using the division result of region growth segmentation;The topological structure is used to generate the house type drawing of the residence.The application combines traditional method and deep learning technology, has strong universality, can adapt to various room structures and complex scene;Through RANSAC fitting technology, the precision of straight line detection is significantly improved, can be used for fine design, improve residence measurement efficiency and home decoration scene User experience.

Description

Technical Field

[0001] This invention relates to a method and system for automatically generating architectural floor plans based on three-dimensional point clouds. Background Technology

[0002] Automatic generation technology for residential floor plans is an important technology in the fields of modern interior design and building information technology (BIM). This technology aims to convert 3D point cloud data (which is usually collected by devices such as RGB-D cameras and laser scanners) into accurate and easy-to-understand 2D floor plans.

[0003] The generated floor plan is presented from a bird's-eye view, representing the interior spatial layout, room divisions, and the location of doors and windows, providing crucial information support for interior design, robot navigation, and augmented reality (AR) / virtual reality (VR) applications. The involved 3D point cloud is a data set composed of a large number of points in 3D space. These points carry information such as spatial coordinates, color, and normals, accurately describing the 3D shape of objects.

[0004] Existing floor plan generation algorithms can be broadly categorized into two types: geometric reconstruction-based methods and deep learning-based methods. Geometric reconstruction-based methods typically rely on accurate 3D reconstruction results, but they perform poorly when dealing with non-Manhattan structures (i.e., non-right-angle layouts). Deep learning-based methods, on the other hand, generate floor plans by learning features from large amounts of data. While they have stronger generalization capabilities, they still suffer from poor performance when handling complex scenarios.

[0005] Existing methods for automatically generating building floor plans cannot accurately segment different room areas from point clouds, nor can they identify and locate key elements such as doors and windows. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to overcome the shortcomings of existing automatic building floor plan generation methods, which cannot accurately segment different room areas from point clouds and cannot identify and locate key elements such as doors and windows. The present invention provides an automatic building floor plan generation method and system based on three-dimensional point clouds that significantly improves the accuracy of line detection, can be used for refined design, and improves the efficiency of residential measurement and the user experience of home decoration scenarios.

[0007] The present invention solves the above-mentioned technical problems through the following technical solution:

[0008] A method for automatically generating architectural floor plans based on 3D point clouds, characterized in that the method includes:

[0009] Obtain 3D point cloud data of a residential building;

[0010] The three-dimensional point cloud data is preprocessed to obtain a residential model;

[0011] The residential model is sliced ​​using projection, with the projection slices perpendicular to the Z-axis.

[0012] Perform region growing and segmentation on the projected slice image;

[0013] The topological structure between the fitted lines in the projected slices is established using the partitioning results of region growing.

[0014] The floor plan of the residence is generated using the aforementioned topology.

[0015] In this application, "residential" refers to a residential unit. The dataset consists of 3D point cloud data collected from typical residential units in Chinese cities. The target is standard apartment units (such as three-bedroom / two-bedroom / one-living room layouts) in high-rise buildings within residential communities. The data comprehensively covers all functional spaces within the residence, including independent zones such as living room, bedrooms × N, kitchen, bathroom, and balcony, as well as transitional areas such as corridors and doorways connecting these spaces. The point cloud data was acquired through multi-station laser scanning, fully recording the building's structural features (partition walls, beams, columns, door and window openings) and spatial scale information, while also including point cloud features of some fixed furniture (built-in cabinets, bathroom facilities).

[0016] Preferably, the projection slicing of the residential model includes:

[0017] Use 3D point cloud data to obtain coordinate units, and filter point cloud data based on coordinate units;

[0018] Iterative calculations are performed on the 3D point cloud data to obtain the maximum and minimum values ​​in the height direction within the height range;

[0019] The residential model is projected and sliced ​​within the range of maximum and minimum values.

[0020] Preferably, the projection slicing of the residential model further includes:

[0021] Binarization of the projected slice image based on millimeter-level resolution;

[0022] Use the MLSD algorithm to extract long line segments from an image;

[0023] Perform cluster analysis on the extracted line segments, extract the median line of a cluster, and calculate the average angle of the line segments;

[0024] The rotation matrix is ​​calculated based on the difference between the median line segment angle and the X-axis. The rotation matrix is ​​then used to rotate the point cloud data into a unified XY coordinate system.

[0025] Preferably, the formula for calculating the rotation matrix is:

[0026]

[0027] Where θ is the rotation angle around the origin.

[0028] The formula for calculating the angle of a line segment is:

[0029]

[0030] in, and These are the starting and ending coordinates of the line segment, respectively. The arctangent function is used to calculate the angle of the line segment. α is used as an adjustment constant. When the calculated angle is positive, its value is zero. When the calculated angle is negative, α is increased by 180 degrees to make it positive.

[0031] Preferably, the region growing segmentation of the projected slice image includes:

[0032] The projected slice image is segmented by region growing to generate a preliminary room segmentation map;

[0033] Obstacles and door frames are simulated by dilation operations. The outline is obtained by dilation based on the door width as a coefficient and then grown. The projected slice image is then divided into regions.

[0034] Rooms are grouped into the same area by using sparse areas of point clouds.

[0035] Preferably, the step of establishing the topological structure between the fitted lines in the projected slices using the partitioning results of region growing segmentation includes:

[0036] Extract the maximum outline of the room from the projected slice image;

[0037] Match the point cloud to line segments for the room boundaries, and transform the boundary points into normalized line segments;

[0038] Cluster line segments and classify and label components to generate the topology of structural components.

[0039] Preferably, the topology of the generated structural component includes:

[0040] Extract lines from the projected slice image and perform linearization processing;

[0041] Merge regions of similar lines;

[0042] Obtain sub-structural components from the merged region;

[0043] Analyze and adjust the topology of the point cloud line segments to perform topology processing on the sub-structure components;

[0044] The projected slice image is subjected to region growing processing based on distance standards;

[0045] The final room partition map is generated based on the region growth process;

[0046] The walls, doors, and windows are clustered and semantically analyzed.

[0047] Preferably, the method for automatically generating architectural floor plans includes:

[0048] The topology of structural components is obtained by using the distance from data points to line segments during the expansion operation.

[0049] Specifically, the topological structure between the fitted lines in the projected slices is established using the partitioning results of region growing and segmentation, including:

[0050] The shape of the outer boundary points of multiple areas filled in the residence is grown from the obtained projected slice image area;

[0051] Long line segments are extracted from the image by using the MLSD algorithm after obtaining the projected slice image;

[0052] Match the outer boundary points and long line segments, arrange the long line segments in a counterclockwise order, and establish the topological relationship between points and lines between the line segments;

[0053] The long line segments of a single region are intersected, merged, and completed according to their sequential relationship, and standardized into right-angled polygons;

[0054] Project the point cloud of walls near the line segment onto the line segment, and mark the gaps as holes. Based on the topological distribution of all region centers and wall line segments, identify the holes as doors and windows, and establish the topological structure of doors, windows, and walls.

[0055] Preferably, the formula for the distance from a data point to a line segment is:

[0056]

[0057] in, The coordinates of the current data point during the dilation operation. These are the coordinates of several points within the line segment;

[0058] The formula for the angle between a data point and a midpoint of a line segment is:

[0059]

[0060] The matching index for line segments is:

[0061]

[0062] When the number of line segments is not zero, it is necessary to calculate the average number of line segments.

[0063] The present invention also provides an automatic floor plan generation system, characterized in that the automatic floor plan generation system is used to implement the automatic generation method of building floor plans based on three-dimensional point clouds as described above.

[0064] Based on common knowledge in the field, the above-mentioned preferred conditions can be combined arbitrarily to obtain various preferred embodiments of the present invention.

[0065] The positive and progressive effects of this invention are as follows:

[0066] This invention can process 3D point cloud data more accurately for room segmentation, and generate a complete room floor plan including the distribution of all doors, windows and walls, which is convenient for construction workers and users and improves construction production efficiency.

[0067] Based on 3D point cloud data, this application uses advanced algorithm models to accurately extract and visualize key building structures (doors, windows, and walls). Attached Figure Description

[0068] Figure 1 This is a flowchart of the automatic generation method for building floor plans according to Embodiment 1 of the present invention.

[0069] Figure 2 This is a rendering of the automatic generation method for building floor plans according to Embodiment 1 of the present invention. Detailed Implementation

[0070] The present invention will be further illustrated by way of embodiments below, but the present invention is not limited to the scope of the embodiments described herein.

[0071] Example 1

[0072] This embodiment provides an automatic floor plan generation system based on 3D point clouds. The automatic floor plan generation system includes a laser scanning device and a processing module. The processing module can be integrated into the laser scanning device or it can be a processing terminal that can receive scanning data from the laser scanning device.

[0073] The processing module is used for:

[0074] Obtain 3D point cloud data of a residential building;

[0075] The three-dimensional point cloud data is preprocessed to obtain a residential model;

[0076] The residential model is sliced ​​using projection, with the projection slices perpendicular to the Z-axis.

[0077] Perform region growing and segmentation on the projected slice image;

[0078] The topological structure between the fitted lines in the projected slices is established using the partitioning results of region growing.

[0079] The floor plan of the residence is generated using the aforementioned topology.

[0080] Furthermore, the processing module is used for:

[0081] Use 3D point cloud data to obtain coordinate units, and filter point cloud data based on coordinate units;

[0082] Iterative calculations are performed on the 3D point cloud data to obtain the maximum and minimum values ​​in the height direction within the height range;

[0083] The residential model is projected and sliced ​​within the range of maximum and minimum values.

[0084] Furthermore, the processing module is used for:

[0085] Binarization of the projected slice image based on millimeter-level resolution;

[0086] Use the MLSD algorithm to extract long line segments from an image;

[0087] Perform cluster analysis on the extracted line segments, extract the median line of a cluster, and calculate the average angle of the line segments;

[0088] The rotation matrix is ​​calculated based on the difference between the median line segment angle and the X-axis. The rotation matrix is ​​then used to rotate the point cloud data into a unified XY coordinate system.

[0089] The formula for calculating the rotation matrix is:

[0090]

[0091] Where θ is the rotation angle around the origin.

[0092] The formula for calculating the angle of a line segment is:

[0093]

[0094] in, and These are the starting and ending coordinates of the line segment, respectively. The arctangent function is used to calculate the angle of the line segment. α is used as an adjustment constant. When the calculated angle is positive, its value is zero. When the calculated angle is negative, α is increased by 180 degrees to make it positive.

[0095] Furthermore, the processing module is used for:

[0096] The projected slice image is segmented by region growing to generate a preliminary room segmentation map;

[0097] Obstacles and door frames are simulated by dilation operations. The outline is obtained by dilation based on the door width as a coefficient and then grown. The projected slice image is then divided into regions.

[0098] Rooms are grouped into the same area by using sparse areas of point clouds.

[0099] Furthermore, the processing module is used for:

[0100] Extract the maximum outline of the room from the projected slice image;

[0101] Match the point cloud to line segments for the room boundaries, and transform the boundary points into normalized line segments;

[0102] Cluster line segments and classify and label components to generate the topology of structural components.

[0103] Furthermore, the processing module is used for:

[0104] Extract lines from the projected slice image and perform linearization processing;

[0105] Merge regions of similar lines;

[0106] Obtain sub-structural components from the merged region;

[0107] Analyze and adjust the topology of the point cloud line segments to perform topology processing on the sub-structure components;

[0108] The projected slice image is subjected to region growing processing based on distance standards;

[0109] The final room partition map is generated based on the region growth process;

[0110] The walls, doors, and windows are clustered and semantically analyzed.

[0111] Furthermore, the processing module is used for:

[0112] The topology of structural components is obtained by using the distance from data points to line segments during the expansion operation.

[0113] Specifically, the processing module is used for:

[0114] The shape of the outer boundary points of multiple areas filled in the residence is grown from the obtained projected slice image area;

[0115] Long line segments are extracted from the image by using the MLSD algorithm after obtaining the projected slice image;

[0116] Match the outer boundary points and long line segments, arrange the long line segments in a counterclockwise order, and establish the topological relationship between points and lines between the line segments;

[0117] The long line segments of a single region are intersected, merged, and completed according to their sequential relationship, and standardized into right-angled polygons;

[0118] Project the point cloud of the wall near the line segment onto the line segment, mark the gaps as holes, and identify the holes as doors and windows based on the topological distribution of all region centers and wall line segments, and establish the topological structure of doors, windows and walls.

[0119] Furthermore, the processing module is used for:

[0120] The formula for the distance from a data point to a line segment is:

[0121]

[0122] in, The coordinates of the current data point during the dilation operation. These are the coordinates of several points within the line segment;

[0123] The formula for the angle between a data point and a midpoint of a line segment is:

[0124]

[0125] The matching index for line segments is:

[0126]

[0127] When the number of line segments is not zero, it is necessary to calculate the average number of line segments.

[0128] The automatic floor plan generation system in this embodiment generates floor plans through the following steps: First, it filters the residential point cloud data by height range and projects multi-layer slices onto the XY plane. Second, it analyzes the point cloud density using a 5mm grid to extract wall point clouds. Third, it employs a lightweight convolutional neural network to extract features from the sliced ​​images, generating pixel-level line segment heatmaps and line segment vector fields to accurately locate line segment regions and regress the direction and length of the line segments. Fourth, it performs rotation transformation on the point cloud data by analyzing the slope of the line segments, making the walls parallel to the XY axes. Fifth, it further performs region growing segmentation and dilation processing on the image to achieve room partitioning. Sixth, it combines the extracted line segment boundary information with the partitioning results to perform line segment matching, closure, and regularization, generating right-angled polygons for each room partition. Finally, it projects the vectorized partition line segments onto the point cloud and analyzes the distance from the line to the point cloud for accurate fitting, identifying all empty areas and marking them as doors and windows. Based on global topological information and the morphological distribution of doors and windows, a complete room topology structure including elements such as walls, doors, and windows is constructed and exported as PNG images, JSON data, and DXF vector graphics files. This invention combines traditional methods with deep learning technology, possessing strong versatility and adaptability to various room structures and complex scenarios. Through RANSAC fitting technology (random sampling consistency), the accuracy of line detection is significantly improved, enabling its use in refined design, enhancing residential measurement efficiency, and improving the user experience in home decoration scenarios.

[0129] See Figure 1 Using the aforementioned floor plan automatic generation system, this embodiment also provides a method for automatically generating architectural floor plans, including:

[0130] Step 100: Obtain 3D point cloud data of a residential building;

[0131] Step 101: Preprocess the three-dimensional point cloud data to obtain a residential model;

[0132] Step 102: Project and slice the residential model;

[0133] Step 103: Perform region growing segmentation on the projected slice image;

[0134] Step 104: Use the partitioning results of region growing to establish the topological structure between the fitted lines in the projected slices;

[0135] Step 105: Generate the floor plan of the residence using the topology.

[0136] In this embodiment, the input 3D point cloud undergoes preprocessing and region segmentation. The input 3D point cloud data includes: multi-point scanning of indoor or outdoor scenes using a LiDAR scanner to obtain high-density point cloud data. The scanning device provides millimeter-level precision to ensure data integrity and accuracy. The scanning results are saved in a standardized point cloud format (e.g., .pts). Data preprocessing operations include: data filtering, iterative calculation, and projection slicing; binarization of the sliced ​​point cloud; calculation of the rotation matrix, and rotation and orientation of the point cloud data according to the rotation matrix; region growing segmentation of the sliced ​​image; generation of the closed topology of the sliced ​​image, and annotation of doors, windows, and walls. The output module is used to generate and output residential floor plans, which can be exported as PNG images, JSON format data, and DXF vector graphics files.

[0137] Step 102 specifically includes:

[0138] 1021. Obtain coordinate units from 3D point cloud data and filter point cloud data based on coordinate units;

[0139] 1022. Perform iterative calculations on the 3D point cloud data to obtain the maximum and minimum values ​​in the height direction within the height range;

[0140] 1023. Project slices the residential model within the range of maximum and minimum values;

[0141] 1024. Binarize the projected slice image based on millimeter-level resolution;

[0142] 1025. Use the MLSD algorithm to extract long line segments from an image;

[0143] 1026. Perform cluster analysis on the extracted line segments, extract the median line of a cluster, and calculate the average angle of the line segments;

[0144] 1027. Calculate the rotation matrix based on the difference between the median line segment angle and the X-axis, and use the rotation matrix to rotate the point cloud data into a unified XY coordinate system.

[0145] In this embodiment, point cloud data is filtered by judging the coordinate unit of the input point cloud; the maximum and minimum values ​​in the z-axis direction of the point cloud data are iteratively calculated to limit the height range of the point cloud to between 2 meters and 5 meters; the point cloud data is projected into multi-layer slices for easy subsequent partitioning processing; the point cloud slices are binarized based on millimeter-level resolution to generate high-precision slice images; the MLSD algorithm (long line detection) is used to extract long line segments in the slices, and for wall points, a lightweight convolutional neural network is used to extract features to predict line segment heatmaps and vector fields; the extracted line segments are clustered to extract the median of each cluster and calculate the average angle of the line segments; a rotation matrix is ​​calculated based on the line segment angles to rotate the point cloud data to a unified XY coordinate system.

[0146] The formula for calculating the rotation matrix is ​​as follows:

[0147]

[0148] Where θ is the rotation angle around the origin. If the selected line angle is closer to the X-axis, θ is the negative value of the calculated line segment angle. If the line segment is closer to the Y-axis, then θ is the difference between the line segment angle and 90 degrees.

[0149] The formula for calculating the angle of a line segment is:

[0150]

[0151] in, and These are the starting and ending coordinates of the line segment, respectively. The arctangent function is used to calculate the angle of the line segment. α is used as an adjustment constant. When the calculated angle is positive, its value is zero. When the calculated angle is negative, α is increased by 180 degrees to make it positive.

[0152] Step 103 specifically includes:

[0153] Step 1031: Perform region growing segmentation on the projected slice image to generate a preliminary room segmentation map;

[0154] Step 1032: Simulate obstacles and door frames through dilation operation, and grow the outline after dilation with the door width as the coefficient, and then divide the projected slice image into regions.

[0155] Step 1033: Group the rooms into the same area by using the sparse areas of the point cloud.

[0156] Step 104 specifically includes:

[0157] Step 1041: Extract the maximum outline of the room from the projected slice image;

[0158] Step 1042: Match the point cloud to line segments for the room boundary, and convert the boundary points into normalized line segments;

[0159] Step 1043: Cluster line segments and classify and label components to generate the topology of structural components.

[0160] Step 1043 specifically includes:

[0161] Extract lines from the projected slice image and perform linearization processing;

[0162] Merge regions of similar lines;

[0163] Obtain sub-structural components from the merged region;

[0164] Analyze and adjust the topology of the point cloud line segments to perform topology processing on the sub-structure components;

[0165] The projected slice image is subjected to region growing processing based on distance standards;

[0166] The final room partition map is generated based on the region growth process;

[0167] The walls, doors, and windows are clustered and semantically analyzed.

[0168] The method for automatically generating architectural floor plans includes:

[0169] The topology of structural components is obtained by using the distance from data points to line segments during the expansion operation.

[0170] Specifically, the topological structure between the fitted lines in the projected slices is established using the partitioning results of region growing and segmentation, including:

[0171] The shape of the outer boundary points of multiple areas filled in the residence is grown from the obtained projected slice image area;

[0172] Long line segments are extracted from the image by using the MLSD algorithm after obtaining the projected slice image;

[0173] Match the outer boundary points and long line segments, arrange the long line segments in a counterclockwise order, and establish the topological relationship between points and lines between the line segments;

[0174] The long line segments of a single region are intersected, merged, and completed according to their sequential relationship, and standardized into right-angled polygons;

[0175] Project the point cloud of walls near the line segment onto the line segment, and mark the gaps as holes. Based on the topological distribution of all region centers and wall line segments, identify the holes as doors and windows, and establish the topological structure of doors, windows, and walls.

[0176] The formula for the distance from a data point to a line segment is:

[0177]

[0178] in, The coordinates of the current data point during the dilation operation. These are the coordinates of several points within the line segment;

[0179] The formula for the angle between a data point and a midpoint of a line segment is:

[0180]

[0181] The matching index for line segments is:

[0182]

[0183] When the number of line segments is not zero, it is necessary to calculate the average number of line segments.

[0184] In this embodiment, region growing segmentation is performed on the sliced ​​image to generate a preliminary room segmentation map; dilation operation is used to simulate obstacles and door frames, naturally dividing the room areas; in the case of double doors or unfinished rooms, rooms are grouped into the same area by sparse regions of the point cloud; the maximum contour of the room is extracted from the sliced ​​image, and line segments are matched to the boundary to form a closed topological structure of the room; point cloud to line matching is performed on the room boundary to convert the boundary points into normalized line segments; line segments are clustered and walls, doors and windows are classified and labeled to generate a complete room topological structure. Lines extracted from the point cloud are linearized; similar point cloud regions are merged; the merged region is further divided into smaller parts; the topological structure of the point cloud line segments is analyzed and adjusted for topological processing; the point cloud is region grown according to the distance standard; the point cloud data is segmented into different parts; walls, doors and windows are clustered and semantically analyzed. Specifically, a room floor plan is generated based on the topological structure, including the distribution of all walls, doors and windows; all wall and door data are output; the processed data is exported in PNG, JSON and DXF formats.

[0185] See Figure 2 (a) shows the preprocessed data effect; (b) shows the room zoning effect; and (c) shows the output result. Specifically, in this embodiment, the color image in (b) is generated using growth segmentation. During the dilation process, the dilated data points are digitized points, and their coordinates and the distances between coordinates are known. The distance between two points is obtained using the dilated data points. Based on this distance, it can be determined whether the current area is a door frame area. Moreover, the semantics of the boundary points of the color image are known. In this embodiment, the semantics of the boundary points and the distances from the dilated data points on the boundary points to the point cloud and line segments on the point cloud slice can be used to obtain the distance and positional relationship between each line segment and the boundary point. The semantics of the image points on the point cloud slice can be obtained using these distances and positional relationships.

[0186] While specific embodiments of the present invention have been described above, those skilled in the art should understand that these are merely illustrative examples, and the scope of protection of the present invention is defined by the appended claims. Those skilled in the art can make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but all such changes and modifications fall within the scope of protection of the present invention.

Claims

1. A method for automatic generation of building floor plans based on three-dimensional point clouds, characterized in that, The building plan automatic generation method comprises: acquiring three-dimensional point cloud data of a residence; preprocessing the three-dimensional point cloud data to obtain a residence model; projecting and slicing the residence model, the projection and slicing being perpendicular to the Z-axis direction; region growing segmentation of the projected and sliced image; establishing a topological structure between straight lines fitted in the projected and sliced image by using the division result of the region growing segmentation; generating a house plan of the residence by using the topological structure; the region growing segmentation of the projected and sliced image comprises: region growing segmentation of the projected and sliced image to generate a preliminary room segmentation map; regionally separating the projected and sliced image by simulating obstacles and door frames through inflation operation and growing after inflation contouring according to a door width coefficient; classifying rooms in the same region through point cloud sparse regions; the establishment of the topological structure between the straight lines fitted in the projected and sliced image by using the division result of the region growing segmentation comprises: extracting a maximum contour of a room from the projected and sliced image; point cloud to line segment matching of a room boundary to convert boundary points into normalized line segments; clustering line segments and classifying and labeling components to generate a topological structure of structural components; the generation of the topological structure of structural components comprises: extracting lines in the projected and sliced image and performing linearization processing; regionally merging similar lines; acquiring sub-structural components after region merging; topological processing of sub-structural components by analyzing and adjusting the topological structure of point cloud line segments; region growing processing of the projected and sliced image according to a distance standard; generating a final room segmentation map according to the region growing processing; clustering and semantically analyzing walls, doors and windows; wherein the establishment of the topological structure between the straight lines fitted in the projected and sliced image by using the division result of the region growing segmentation comprises: obtaining shape outer boundary points of a plurality of regions filled in the residence by region growing from the projected and sliced image; extracting long line segments in the picture by using an MLSD algorithm from the projected and sliced image; matching the outer boundary points and the long line segments, arranging the long line segments in a counterclockwise order, and establishing a point and line topological relationship between the line segments; intersecting, merging and completing the long line segments of a single region according to the order relationship to normalize the long line segments into right-angle polygons; projecting wall point clouds near the line segments to the line segments, marking vacancies as voids, and identifying the voids as doors and windows according to the topological distribution of all region centers and wall line segments to establish a topological structure of the doors, the windows and the walls. 2.The three-dimensional point cloud based building plan automatic generation method of claim 1, wherein, the projection and slicing of the residence model comprises: obtaining a coordinate unit by using the three-dimensional point cloud data, and screening point cloud data according to the coordinate unit; iteratively calculating the three-dimensional point cloud data to obtain a maximum value and a minimum value in a height direction within a height range; projecting and slicing the residence model within an interval of the maximum value and the minimum value. 3.The method of claim 2, wherein, the projection and slicing of the residence model further comprises: performing binaryzation processing on the projected and sliced image based on millimeter-level resolution; extracting long line segments in the picture by using an MLSD algorithm; performing clustering analysis on the extracted line segments to extract a median line of a cluster; calculating a rotation matrix according to a difference between an angle of the median line segment and an X-axis, and rotating point cloud data to a unified XY coordinate system by using the rotation matrix. 4.The method of claim 3, wherein, The formula for calculating the rotation matrix is: ， Wherein, theta is the rotation angle around the origin, The formula for calculating the angle of the line segment is: ， wherein, and are the start and end point coordinates of the line segment, the arctangent function is used to calculate the angle of the line segment, and α is an adjustment constant. When the calculated angle is positive, the value is zero, and when the calculated angle is negative, it is changed to positive by adding 180 degrees to α. 5.The method of claim 1, wherein, The topology of the structure is obtained by using the distance between the data points and the line segment in the expansion operation, and the formula for the distance between the data points and the line segment is: ， wherein, the coordinates of the current data point in the expansion operation, are the coordinates of the number points in the line segment; The formula for the angle between the data points and the points in the line segment is: ， The matching index of the line segment is: ， Wherein, when the number of line segments is not zero, the average value of the number of line segments needs to be calculated.

6. A house type drawing automatic generation system, characterized in that, The house type diagram automatic generation system is used to realize the three-dimensional point cloud based building plan automatic generation method as claimed in any one of claims 1 to 5.

Images