A method for generating a three-dimensional house model by using a house surveying plan

Abstract

The present application relates to the technical field of housing information processing, in particular to a kind of housing surveying planar map generation three-dimensional house model analysis method, comprising: based on housing surveying planar map, record the spatial data of three-dimensional house model under multiple housing structures;Based on the spatial data of housing structure, cross validation is carried out with the wall in the position of housing structure, and the difference classification of each wall and standard model is marked;With the difference classification corresponding to each wall, construct random change matrix, determine the boundary feature of each wall in horizontal position and vertical position;With the coordinate corresponding to boundary feature as input, the geometry of wall is verified, and under the condition that geometry verification is established, the coordinate point set of each wall is determined;With the difference classification and boundary feature of coordinate point set as basis, re-encode three-dimensional house model, and the housing structure after re-encoding is used as the output three-dimensional house model. The accuracy and efficiency of housing model are improved.

Description

Technical Field

[0001] This invention relates to the field of housing information processing technology, specifically a method for generating three-dimensional floor plan models from house surveying and mapping floor plans. Background Technology

[0002] Most properties currently listed for sale online have vectorized floor plan data, which contains a wealth of floor plan features. A floor plan consists of rooms, each typically an enclosed area bounded by walls. The displayed area usually represents the layout of the walls, doors, and windows within the room. When describing floor plans using house surveying maps, only the dimensions of doors and windows and the overall room size are typically emphasized, often neglecting wall widths. This can lead to a distorted perception of room size when users view properties online, reducing the effective display rate of data and potentially causing user churn.

[0003] For example, Chinese Patent Publication No. CN114791964A discloses a method, apparatus, electronic device, and storage medium for processing home decoration videos. The method includes: when a user browses the decoration effect of a house online, the terminal can generate model adjustment information for adjusting the home decoration model based on the position information of the home decoration model in the three-dimensional house space and the position information of the target roaming point in the three-dimensional house space during the loading of home decoration data. Then, the initial keyframes corresponding to the home decoration model are adjusted to generate target keyframes. The target keyframes can then be played before the data loading is completed. On the one hand, playing the corresponding home decoration video can reduce the user's bad experience while waiting for data loading, prevent users from exiting the browsing due to excessive waiting time, and improve the effective display rate of decoration data.

[0004] For example, Chinese Patent Publication No. CN111339336A discloses a method and system for parsing house layout data, belonging to the field of housing information processing technology. The method includes: obtaining a house layout coordinate set based on house layout data containing wall size information and window size information; determining the metadata type of the house layout based on the house layout coordinate set; determining filtering and judgment conditions based on the metadata type; filtering the house layout coordinate set according to the filtering conditions to obtain a key geometric point coordinate set; and determining the metadata of the house layout corresponding to the key geometric point coordinate set based on the metadata type determined by the house layout coordinate set and the judgment conditions.

[0005] Existing technologies use room panning and scaling to illustrate specific room layout data, and use protrusion recognition to display protrusions and bay windows in the house. These methods emphasize the layout of the room, but ignore the fact that missing annotations on the house survey map can cause discrepancies in the displayed room size, leading to misidentification of the size of the space by the user. Summary of the Invention

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for generating a three-dimensional house type model from a house surveying floor plan, comprising: S1, constructing the house structure of the three-dimensional house type model based on the house surveying floor plan, and recording the spatial data of the three-dimensional house type model under multiple house structures according to the scene type.

[0007] S2, based on the spatial data of the building structure, performs cross-validation on the walls at the location of the building structure, and marks the differences between each wall and the standard model.

[0008] S3. Based on the differences in each wall, construct a random variation matrix to determine the boundary features of each wall in the horizontal and vertical positions.

[0009] S4 takes the coordinates corresponding to the boundary features as input to perform geometric verification on the wall. If the geometric verification is successful, the set of coordinate points of each wall is determined.

[0010] S5 uses the difference classification of coordinate point sets and boundary features as a basis to recode the 3D house model, and uses the recoded house structure as the output 3D house model.

[0011] The beneficial effects of this invention are as follows: First, this invention extracts geometric elements from a house survey plan and uses the minimum bounding box algorithm to calculate the three-dimensional spatial position and dimensions of each structure, combining this with topological sorting to generate a list of house structures. Through spatial positional relationship derivation and regulatory constraints, it ensures that the three-dimensional model conforms to building codes in terms of geometric scale and topological connections, forming an initial three-dimensional structure with spatial logic, providing a benchmark for subsequent difference analysis.

[0012] Second, this invention is based on the Euclidean distance deviation between the measured coordinates and theoretical coordinates of the wall. It fits the deviation points into a difference plane through spatial clustering. It calls a standard wall shape library and compares the tilt angle and center offset to achieve objective quantitative classification of the difference type, providing a reliable basis for the construction of random matrices.

[0013] Third, this invention classifies differences into rows and horizontal / vertical deviations into columns, matching normal / uniform / log-normal distributions to generate random variation matrices; it filters constraints through deviation probabilities to determine reference constraint thresholds for each difference category; then it retrieves wall inflection points and plane midpoints, obtains boundary features through angle alignment and coordinate verification, integrates difference information and boundary features, and generates a 3D model that conforms to building codes; this improves the accuracy and efficiency of generating 3D models from building survey floor plans. Attached Figure Description

[0014] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0015] Figure 1 This is a flowchart illustrating a method for generating a 3D floor plan model from a house surveying and mapping plan.

[0016] Figure 2 This is a flowchart illustrating step S1 of a method for generating a 3D floor plan analysis from a house survey floor plan.

[0017] Figure 3 This is a flowchart illustrating step S2 of a method for generating a 3D floor plan analysis from a house survey floor plan.

[0018] Figure 4 This is a flowchart illustrating step S3 of a method for generating a 3D floor plan analysis from a house survey floor plan. Detailed Implementation

[0019] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. Where specific techniques or conditions are not specified in the embodiments, they shall be performed in accordance with the techniques or conditions described in the literature in the art or in accordance with the product manual.

[0020] See Figure 1 A method for generating a 3D house type model from a house survey floor plan includes: S1, constructing the house structure of the 3D house type model based on the house survey floor plan, and recording the spatial data of the 3D house type model under multiple house structures according to the scene type.

[0021] S2, based on the spatial data of the building structure, performs cross-validation on the walls at the location of the building structure, and marks the differences between each wall and the standard model.

[0022] S3. Based on the differences in each wall, construct a random variation matrix to determine the boundary features of each wall in the horizontal and vertical positions.

[0023] S4 takes the coordinates corresponding to the boundary features as input to perform geometric verification on the wall. If the geometric verification is successful, the set of coordinate points of each wall is determined.

[0024] S5 uses the difference classification of coordinate point sets and boundary features as a basis to recode the 3D house model, and uses the recoded house structure as the output 3D house model.

[0025] When a normal 3D model is generated based on a house survey plan, the current house structure is displayed through a 3D model based on the segmented distances, vertical heights, component heights, survey points, and the length and width of local boundary positions.

[0026] In step S1, it is necessary to construct the standard style of walls, house plan and basic structural style of corresponding house type based on the values ​​marked on the house survey plan. Then, it is necessary to compare and analyze the generated structural styles, and focus on clarifying the detailed parameters (such as wall width and other specific values) that are not marked on the house survey plan, so as to avoid the deviation of area calculation or visual perception error caused by the lack of dimension annotation when previewing the house model online.

[0027] The above-mentioned scenario types distinguish the core uses of houses, such as residential (single-level houses, duplex houses, villa houses, etc.), commercial (offices, shops, hotel rooms, etc.), and industrial (factories, warehouses, etc.). The purpose of classifying and defining houses is to provide targeted guidance for the construction of 3D house models and the recording of spatial data.

[0028] The aforementioned house structure refers to the spatial construction system and component layout of a house, specifically including the three-dimensional layout and connection methods of building components such as walls, doors and windows, floor slabs / beams and columns (the spatial location of load-bearing structures). After explaining the relative positions of these structures in different functional spaces such as rooms, corridors, and balconies, we arrive at the house structure described above.

[0029] like Figure 2 As shown, the implementation of step S1 also includes: S11, generating the house structure based on the house survey floor plan and determining the spatial relationship of the house structure.

[0030] S12. Perform topological sorting based on the spatial relationship of the building structures to obtain a list of building structures, and label each building structure using the dimensions measured on the building survey plan.

[0031] In step S1, after synchronizing the dimensions of all segment distances, vertical heights, component heights, survey points, and local boundary positions, the structure (walls / doors / windows / rooms) is identified. Then, using 3D modeling tools such as Revit and SketchUp, the dimensions annotated on the current house survey plan are imported to generate rectangular wall entities and planar boundaries that distinguish rooms, corridors, balconies, etc., thus obtaining the initial shape of each room. This house model may contain unannotated details such as wall thickness and hidden structural dimensions, which need to be verified and processed in subsequent steps to complete the standard constraints and spatial logic derivation. Finally, the dimensions of the house model are adjusted to avoid errors caused by missing dimensions.

[0032] S13, standardize and constrain the structure of each building to determine the building structure after spatial logical derivation.

[0033] Normative constraints are achieved by comparing the dimensions recorded on the current building survey plan with parameters such as net height and minimum interior wall thickness specifically marked in documents such as the "Unified Standard for Civil Building Design". When the corresponding normative documents are met, the current surveyed dimensions are considered to comply with the normative constraints. For dimensions such as walls that are not directly marked, they are set based on the conventional construction parameters of similar house types. For example, the commonly used interior wall thicknesses are 120mm and 200mm. An intermediate value can be selected as a temporary substitute and then verified and adjusted later to obtain the initial building structure after normative constraints.

[0034] The spatial dimension logical derivation uses the overall dimensions already marked on the survey map (such as the total width and height of the room), subtracts the occupied dimensions of the known components, and calculates the unmarked parameters in reverse to complete the unmarked parameters. Finally, it checks the spatial connection relationships such as the connection between the wall and the floor slab, and the embedding of doors and windows into the wall. When the current house structure is normally connected, its dimensions and coordinates are recorded and used as the output spatial data.

[0035] The final output spatial data will include the specific dimensions and coordinates of the corresponding building structure. This data will be stored as a structured file to facilitate subsequent cross-validation and boundary feature extraction.

[0036] Preferably, in step S11, when generating the house structure based on the house survey plan, the implementation method further includes: classifying geometric elements based on the measurement data of the house survey plan, and clarifying the spatial relationships of each geometric element; geometric elements can be described as main elements and auxiliary elements. Main elements are generally obvious structures in the house survey such as wall lines, door and window openings, columns, and stairs, while auxiliary elements are specific textual annotations such as dimension markings, elevation symbols, and text annotations; spatial relationships describe the room outlines and connectivity relationships. The room outlines can be directly converted from the house survey plan, and the connectivity relationships are obtained through connectivity analysis, which describes the connected components existing in the corresponding structure to obtain the corresponding house structure.

[0037] Preferably, the specific methods for classifying geometric elements are as follows: ① Distinguish by CAD drawing layers, such as wall lines located in the wall layer, columns located in the structural column layer, and dimension annotations located in the dimension layer; ② Identify by graphic features, such as wall lines being continuous straight lines, door and window openings being interrupted areas on the wall lines, and dimension annotations being a combination of numbers and arrows, with the arrows pointing to the corresponding main elements; ③ Associate by text annotations, such as the annotation of the thickness of the inner wall being directly associated with the wall line element.

[0038] Connectivity analysis here describes the boundaries generated by the building structures, determines the locations where the corresponding building structures are connected, and treats these locations as connected relationships.

[0039] For each geometric element, the minimum bounding box calculation is performed on the corresponding house structure to obtain the spatial location and size information of each house structure. Combined with the classification of the geometric elements of the house structure, the spatial location relationship of the house structure is used.

[0040] The minimum bounding box is obtained by extracting the coordinates of the building structure on the boundary. The spatial coordinates of the building structure are described by the boundary point set, and these values ​​are combined to form the output spatial position relationship.

[0041] In one embodiment of the present invention, cross-validation further includes difference state identification to further determine the relative state of each wall in the current building structure. The current solution can use methods such as grid positioning and irregular arc marking. The distance deviation of each wall in the building structure is calculated, and when a deviation occurs according to the distance deviation description, the corresponding wall is classified as a standard wall, an irregular wall, and the functional attribute classification of the corresponding wall. The content of these data combinations is used as the output difference classification, and finally it is determined whether the wall belongs to multiple forms of difference such as translation, depression, protrusion, or curved irregular shape.

[0042] It should be noted that translation here means that there is an error in the plane of the entire wall, and that the width of a wall is off. Depression and bulge mean that the current wall is not a pure rectangle, and that there is a coordinate deviation in some parts of the irregular wall, which is reflected in the 3D model as depression and bulge. Curved irregularity means that the points with coordinate differences are connected to form a curved surface. All of these represent the forms of coordinate differences.

[0043] like Figure 3As shown, the implementation method of step S2 includes: S21, based on the functional attribute classification of the current wall, read the theoretical coordinates of the wall, compare the theoretical coordinates of the wall with the measured coordinates, calculate the Euclidean distance deviation, and obtain the deviation point corresponding to the Euclidean distance deviation; at this time, a standard model of the current house after surveying is established through the BIM model (Building Information Model), and the model is compared with the three-dimensional model generated using the house surveying plan to determine whether there are walls that have been translated or offset in some positions, so as to obtain a three-dimensional illustration containing the corresponding coordinates.

[0044] S22: Spatial clustering is performed on the deviation points. When the clustered deviation points belong to the same wall, the corresponding deviation points are fitted to obtain the difference plane corresponding to the fitted deviation points. When combining features, multiple deviation points are clustered and fitted to form a relatively complete plane, emphasizing the translation and offset forms of the same wall after spatial coordinate combination.

[0045] S23. For deviation points of the same wall, the standard wall shape library is called to compare the tilt angle and center offset of the difference plane with the standard wall shape to obtain the difference classification of each wall. Here, the wall shapes that show differences will be further verified, such as the right angle characteristics and regular size range of rectangular walls, and the deviations they show will be identified, emphasizing the differences in the shapes corresponding to concave, convex and curved irregularities.

[0046] Irregularly shaped walls refer to walls that deviate from the traditional rectangular or straight-line shape, such as curved surfaces (hyperboloids, freeform surfaces), sloping surfaces, polygons, and irregular concave and convex structures.

[0047] Functional attribute classification: According to location, it is divided into exterior walls (enclosure / load-bearing), interior walls (division / filling), gable walls (transverse walls), longitudinal walls, etc.; according to structure, it is divided into solid walls, hollow walls (such as hollow blocks), composite walls (multi-layer composite materials); according to construction method, it is divided into stacked walls (brick / stone), slab walls (cast-in-place concrete), and panel walls (precast wall panels).

[0048] Preferably, the aforementioned center offset represents the Euclidean distance deviation generated by the fitted difference plane at the center of the plane, and the tilt angle represents the tilt angle between the difference plane and the standard model. If the current difference plane is large relative to the standard model, the tilt angle between the difference plane and the standard model will be set by the angle between the normal vectors.

[0049] In step S22, when performing spatial clustering of deviation points, the implementation method also includes: determining whether the Euclidean distance deviation of the current deviation point is a valid deviation. If it is a valid deviation, clustering is performed on each group of deviation points to obtain the clustered deviation points. For example, the deviation points of the east wall and the deviation points of the north wall are clustered separately to avoid confusion of deviation points of different walls.

[0050] For the deviation points after clustering, calculate the spatial distance and normal vector direction between each deviation point. When the normal vector direction is consistent and the spatial distance conforms to the logic of the wall length and width, the corresponding deviation points are regarded as the same wall.

[0051] If the wall is planar, the least squares method is used to fit and generate the difference plane. If the wall is an irregular curved surface (such as a hyperboloid), NURBS surface fitting is used instead of planar fitting to ensure that the difference shape is consistent with the actual shape, so as to obtain the output difference plane. When performing least squares fitting, the fitting minimizes the distance from all deviation points to the plane. For example, if the wall surface is planar, the formula ax + by + cz + d = 0 is used, where (x, y, z) represents the coordinates of the deviation point, (a, b, c) will represent the normal vector of the corresponding deviation point, and d is a constant term that determines the spatial position of the plane, and its value will determine the translation in the direction of the normal vector. When judging the consistency of the normal vectors, the normal vectors will be obtained through multiple deviation points, and then the angle between the calculated normal vectors will be compared to see if it is less than 15 degrees. If it is less than 15 degrees, it is considered that they are consistent. For a continuous plane, treat it as a wall. If the normal vector direction is disordered, it is easy to have some points biased towards the interior and some towards the exterior. It is necessary to check the dimensions marked on the current building survey plan, which may be due to the wrong target point selection during measurement. As for the wall surface being curved, it is necessary to use point cloud fitting to fit it into a curved surface and then calculate its normal vector. As for the spatial distance conforming to the wall length and width logic, it means that the distance between the current deviation points is consistent with the distance between the actual sampling positions. When they are consistent, it is considered to conform to the wall length and width logic.

[0052] If the deviation is not valid, the corresponding deviation point is discarded to eliminate instrument precision error. It should be noted that the determination of valid error at this point is based on the functional attribute classification of the wall structure corresponding to each deviation point. When belonging to different functional attribute classifications, the allowable deviation size for the corresponding wall structure also varies. The allowable error for each type of wall can be retrieved from historical data or specification documents, and the corresponding allowable error can be used as the value for determining the valid deviation at this time. Deviation points greater than this allowable error are then selected as the set of deviation points for the current primary analysis.

[0053] Preferably, when performing spatial clustering of deviation points, clustering is performed according to the spatial location of each deviation point, that is, according to the location marked by each wall, the identified deviation points are classified into a category according to the wall to which they originally belong.

[0054] Preferably, spatial clustering uses the DBSCAN density clustering algorithm: ① The neighborhood radius is set to 50mm, which is magnified 10 times to cover the deviation range based on the conventional accuracy of wall surveying ±5mm; ② The minimum number of samples is set to 3 to ensure that deviation points form a continuous area and avoid interference from single noise points.

[0055] In the general processing of step S23, data such as center offset and tilt angle will be compared with standard walls one by one, and the corresponding difference type will be checked according to their values; at the same time, the plane composed of the corresponding offset points will be used to match the contour shape to further illustrate the difference classification.

[0056] The implementation of step S23 also includes: tracking the difference plane, and when exhaustively searching to the current difference plane, checking the shape of the difference plane to determine the correctness of the shape of the difference plane during exhaustive search.

[0057] If the shape of the current difference plane is incorrectly identified, the exhaustive order of the difference plane exhaustive process is analyzed, and the structural overlap between the corresponding wall and the standard model is determined. The shape of the standard model with the maximum structural overlap is taken as the shape of the current difference plane.

[0058] When the shape of the current difference plane is correctly identified, the corresponding difference plane is output according to the ratio of the tilt angle and center offset of the difference plane.

[0059] During the tracking process, a unique ID is assigned to each discrepancy plane, and then the corresponding wall number, location (e.g., "East Wall - Living Room Section"), and standard model (e.g., "Standard Rectangular Wall, 3000mm long, 2800mm high") are labeled. Then, the contour coordinates, fitting parameters, and list of deviation points of the discrepancy plane are associated.

[0060] Then, based on the probability of the standard shape appearing in similar walls, the high-probability shapes are enumerated first. For example, the enumeration order of the standard model corresponding to the current difference plane is rectangle (probability 85%) → trapezoid (probability 10%) → irregular polygon (probability 5%). For walls with irregular shapes, the enumeration order is single curved surface (probability 60%) → hyperboloid (probability 30%) → irregular curved surface (probability 10%). At this time, by identifying the shape of the current difference plane, the number of sides, interior angles, parallelism of opposite sides, and contour dimensions are determined for the planar shape of the difference plane. For the curved surface shape of the difference plane, the curvature type (single / hyperboloid), curvature value (such as curvature radius in the X / Y direction), and surface boundary smoothness are determined. When the difference between these values ​​and the standard model is less than ±5%, the shape of the corresponding difference plane is considered to be correctly identified. The corresponding tilt angle and center offset are compared with the overall ratio as the output data.

[0061] If any difference is greater than ±5%, it is considered a shape recognition error, and the process is gradually exhaustive. At this point, the similarity between the difference plane and the standard model will be used for exhaustive search, that is, the overlap between the corresponding plane and the standard model will be used to determine its shape, and then the relevant deviation ratio will be output again.

[0062] At this point, the output difference classification will include the corresponding deviation ratio, as well as the corresponding classification of the shape, tilt angle and center offset of the current difference plane in the standard model. The corresponding classification is regarded as the current output difference classification. Each output difference classification will represent a situation in which the wall has deviation.

[0063] In one embodiment of the present invention, in step S3, the difference classification and quantization values ​​of each wall are introduced, as well as the boundaries of the corresponding wall in the horizontal and vertical positions. Then, these data are used to quantify and simulate the dynamic deviation of various difference classifications. Since the difference classification represents the difference plane extracted from each wall, the boundary features of the horizontal and vertical positions will represent the specific boundaries of the corresponding difference plane at the corresponding vertical and horizontal positions. The set random change matrix uses the position index of the wall as the row and the boundary features at the horizontal and vertical positions as the column, and adopts a probability distribution form to illustrate the value of the corresponding deviation at the horizontal and vertical positions. The generated random change matrix will introduce 100 sets of data to cover different deviation scenarios, and make each set of matrices correspond to one deviation case to illustrate the boundary features of each wall under the corresponding difference classification.

[0064] like Figure 4 As shown, the implementation of step S3 includes: S31, using the walls corresponding to each difference category, each wall is arranged with its position index as the row and the deviation dimension of each wall in the horizontal and vertical positions as the column, and multiple sets of random change matrices are generated according to the difference category matching probability distribution.

[0065] S32, based on the deviation probability of the random change matrix, collect the constraint conditions of the wall under different difference classifications, and determine the reference constraint conditions under different difference classifications.

[0066] S33, apply the reference constraint conditions, and use the deviation probability of the corresponding difference classification in the horizontal and vertical positions as the boundary features of the output.

[0067] Preferably, step S31 is implemented by: analyzing the difference classification of the current input, and classifying each wall into different processing methods based on the offset and tilt angle of the difference classification.

[0068] Determine if the current wall belongs to any one of the following categories: offset, tilt, or deformation. Generate the corresponding processing method and, based on the processing method corresponding to the current wall, obtain the deviation probability value for each wall in the random change matrix.

[0069] Preferably, the matching probability distribution will adopt a Gaussian distribution. First, it is determined whether the wall described under the current difference category belongs to any one of the following categories: offset difference, tilt difference, or deformation difference. Offset difference represents a situation where the wall is mainly offset with a small tilt angle. A normal distribution will be constructed using the average offset and standard deviation of the offset under the current difference category. Tilt difference represents a situation where the overall tilt angle is large and the offset is small. A uniform distribution will be used to calculate the probability value. In this case, the minimum and maximum values ​​of the tilt angle will be used to construct a uniform distribution. Deformation difference represents a situation where both the tilt angle and the offset are large. A log-normal distribution will be used. A log-normal distribution will be constructed using the average offset and standard deviation of the offset to obtain the probability value of the wall matching under each difference category.

[0070] It should be noted that the offset used here is different from the center offset used in step S2. At this time, the wall position index emphasizes a single point, and the offset is the offset of the corresponding point, to further explain the offset at the lower boundary of the corresponding center offset.

[0071] Preferably, the aforementioned deviation dimension represents the probability value obtained by the current wall in a horizontal or vertical position, and its value and the description form of the difference classification will represent the deviation dimension at the corresponding position.

[0072] Preferably, when it is determined that the difference belongs to any one of the following categories: offset difference, tilt difference, and deformed difference, a confidence interval is set based on the deviation and tilt angle in the historical data to obtain the offset threshold and tilt angle threshold. The offset threshold is three times the standard deviation plus the average value of the offset at the horizontal or vertical position, and the tilt angle threshold is set based on the average value plus three standard deviations of the historical data. If the offset in the current difference category is greater than the offset threshold and the tilt angle is less than the tilt angle threshold, it is considered an offset difference. If the offset in the current difference category is less than the offset threshold and the tilt angle is greater than the tilt angle threshold, it is considered a tilt difference. Other cases are considered deformed differences.

[0073] Preferably, when collecting constraint conditions in step S32, the implementation method further includes: recording the constraint conditions corresponding to each wall in the way of wall-deviation dimension, the constraint conditions representing the engineering specifications of offset and tilt angle, using deviation probability to filter the constraint conditions, and using the filtered constraint conditions as reference constraint conditions.

[0074] The engineering specifications will be extracted based on the normative documents set during the current model analysis, and the extracted coordinates will be compared to see if they conform to the normal specifications.

[0075] When selecting constraints, firstly, based on the current wall-deviation dimension, multiple deviation points are sorted from highest to lowest deviation probability. The similarity between each sorted deviation point and the constraint is calculated. During similarity calculation, the offset and tilt angle at horizontal / vertical positions are calculated progressively, and their similarity is calculated against the constraint. The average similarity calculated for each point is then used as the similarity of the corresponding deviation point. Next, all deviation points are sorted from highest to lowest similarity, and the top 75% of the sorted data are selected as the current selected constraints. The engineering specifications for the offset and tilt angle represented by these constraints will conform to most scenarios and meet the needs of the current iteration.

[0076] It should be noted that the similarity can be calculated using either cosine similarity or Pearson correlation coefficient. After preprocessing all deviation points, the corresponding similarity can be calculated.

[0077] In one embodiment of the present invention, step S4 is essentially a conversion verification step from feature parameters to spatial coordinates. By verifying geometric rules, it is ensured that the digital representation of the wall conforms to the real constraints of physical space and meets the mathematical requirements of subsequent model reconstruction. Finally, an accurate set of coordinate points is output to support the accurate reconstruction of the three-dimensional house model.

[0078] At this point, geometric verification is a crucial step in ensuring the accuracy of the model. It measures the accuracy of the coordinate points selected from the boundary features, determining whether they conform to the closure, intersection, and connection conditions in spatial geometry. Geometric verification also identifies the boundary features extracted from multiple sets of data, checking whether these boundary features satisfy the requirements of clear difference classification and whether the multiple deviation points obtained can help identify the problem areas of the current model under the corresponding probabilities. Ultimately, it finds multiple sets of coordinates that can assist in modifying the model, thereby improving the accuracy of the house surveying floor plan generation model.

[0079] The implementation of step S4 also includes: retrieving the inflection points of the wall and the midpoints on the wall plane as the set of coordinate points to be processed; aligning multiple sets of points in the coordinate set according to the included angle and verifying their coordinates; and after any verification condition is met, using the set of coordinate points corresponding to the verification condition as the current output data.

[0080] The midpoint on the wall plane is a point on the wall surface between two inflection points. This midpoint and inflection point are used to further verify the geometric structure of the current house surveying and mapping model, and to verify whether the multiple points extracted through difference classification and boundary features are valid coordinates, so as to facilitate subsequent adjustments to the generated 3D model.

[0081] It should be noted that the verification conditions at this time include spatial geometric rule verification, coordinate system consistency verification, topological relationship constraints, and difference point classification logic verification. After any one of the verifications is satisfied, the corresponding data will be used as the point set for subsequent auxiliary correction of the model.

[0082] Verification of spatial geometric rules includes: closure verification, non-crossing verification, and connection rationality verification.

[0083] Closure check: The wall must form a closed polygon (such as a quadrilateral) without openings or breaks. For example, the Polygon class in the Shapely library (a Python code library used to calculate quadrilateral intersections and unions) can be used to check whether the wall boundaries are self-intersecting or unclosed.

[0084] No crossover check: Any two walls have no spatial overlap or intersection. For example, the intersects method (an algorithm for detecting intersections in rectangular regions) is used to detect whether wall boundary segments intersect.

[0085] Connection rationality verification: The coordinates of the endpoints of adjacent walls must be precisely matched (error ≤ 1mm) to ensure structural continuity. For example, the coordinate difference between the right endpoint of wall A and the left endpoint of wall B must be less than 0.001.

[0086] Coordinate system consistency verification includes coordinate system alignment and angle alignment.

[0087] Coordinate system alignment: The coordinate points of all walls must be based on the same global coordinate system to avoid offsets caused by local coordinate systems.

[0088] Angle alignment: Coordinate points are sorted and aligned according to their included angles to ensure that geometric features (such as corners and turning points) maintain a consistent orientation in space. For example, the arctan2 function (a function for calculating angles) in NumPy (an open-source scientific computing library for Python) is used to calculate the angle between the coordinate points and the origin, and then sorted by angle.

[0089] Topological relationship constraints include: shared wall verification, adjacency relationship verification, and hierarchical relationship verification.

[0090] Shared wall verification: The coordinates of the shared wall must be completely coincident (e.g., the coordinates of the shared wall endpoints of wall A and wall B are consistent).

[0091] Adjacent relationship verification: The connection points of adjacent walls must satisfy topological continuity (e.g., the end point of wall A coincides with the start point of wall B).

[0092] Hierarchical relationship verification: The inclusion relationship between walls (such as the main wall and the attached wall) needs to be verified through coordinate nesting relationship verification.

[0093] The difference point classification logic verification differs from the above verification. It uses the currently input coordinates and deviation probability values ​​to invert the relevant coordinates, check whether the deviation probability of each coordinate is accurate, and whether the labels corresponding to the differences such as offset, tilt, and deformation are correct. It also checks the offset and tilt angle, and re-identifies the difference point classification. Finally, the set of coordinate points that have completed arbitrary verification is used as the data basis for subsequent adjustments.

[0094] Step S5, based on the coordinate point set input in Step S4, uses the points in the coordinate point set as a reference to reposition the spatial positions of each wall, adjusts its shape by combining boundary features, and constructs a recoded model framework. Then, difference classification annotations are added to the reconstructed model, and boundary features are labeled to the coordinate points, allowing the model to simultaneously carry both morphological and data information. Finally, the connections such as wall joints in the recoded model are checked to ensure there are no morphological conflicts or data omissions. The checked model is then used as the output 3D apartment model, along with corresponding point annotation documents, to complete the analysis and processing of the 3D apartment model.

[0095] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered within the protection scope of the present invention.

Claims

1. A method for analyzing a three-dimensional house model generated from a housing survey plan, characterized by, include: S1, based on the house survey floor plan, construct the house structure of the 3D house model, and record the spatial data of the 3D house model under multiple house structures according to the scene type; S2, based on the spatial data of the building structure, cross-validates the walls at the location of the building structure and marks the differences between each wall and the standard model; S3, based on the differences corresponding to each wall, construct a random variation matrix to determine the boundary features of each wall in the horizontal and vertical positions; The implementation of step S3 includes: S31, using the walls corresponding to each difference category, dividing each wall into rows with its position index and columns with the deviation dimensions of each wall in the horizontal and vertical positions, and generating multiple sets of random variation matrices according to the difference category matching probability distribution; S32, based on the deviation probability of the random variation matrix, collecting constraint conditions for the walls under different difference categories, and determining the reference constraint conditions under different difference categories; S33, executing the reference constraint conditions, and using the deviation probability of the corresponding difference category in the horizontal and vertical positions as the output boundary features; S4, take the coordinates corresponding to the boundary features as input, perform geometric verification on the wall, and determine the set of coordinate points of each wall if the geometric verification is successful; S5 uses the difference classification of coordinate point sets and boundary features as a basis to recode the 3D house model, and uses the recoded house structure as the output 3D house model.

2. The method of claim 1, wherein the method further comprises: The implementation of step S1 also includes: S11, Generate the building structure based on the building survey floor plan, and determine the spatial relationship of the building structure; S12, perform topological sorting based on the spatial relationship of the building structures to obtain a list of building structures, and label each building structure using the dimensions measured on the building survey plan; S13, standardize and constrain the structure of each building to determine the building structure after spatial logical derivation.

3. The method for generating a three-dimensional floor plan model from a house surveying floor plan according to claim 2, characterized in that, In step S11, when generating the house structure based on the house survey floor plan, the implementation method also includes: Based on the measurement data of the building's floor plan, geometric elements are classified to clarify the spatial relationships between them; For each geometric element, the minimum bounding box calculation is performed on the corresponding house structure to obtain the spatial location and size information of each house structure. Combined with the classification of the geometric elements of the house structure, the spatial location relationship of the house structure is used.

4. The method for generating a three-dimensional floor plan model from a house surveying floor plan according to claim 1, characterized in that, Step S2 can be implemented in the following ways: S21. Based on the functional attribute classification of the current wall, read the theoretical coordinates of the wall, compare the theoretical coordinates of the wall with the measured coordinates, calculate the Euclidean distance deviation, and obtain the deviation point corresponding to the Euclidean distance deviation. S22, perform spatial clustering on the deviation points. When the clustered deviation points belong to the same wall, fit the corresponding deviation points to obtain the difference plane corresponding to the fitted deviation points. S23. For deviation points of the same wall, call the standard wall shape library, compare the tilt angle and center offset of the difference plane with the standard wall shape, and obtain the difference classification of each wall.

5. The method for generating a three-dimensional floor plan model from a house surveying floor plan according to claim 4, characterized in that, When performing spatial clustering on the deviation points in step S22, the implementation method also includes: Determine whether the Euclidean distance deviation of the current deviation point is a valid deviation. If it is a valid deviation, cluster each group of deviation points to obtain the clustered deviation points. For the deviation points after clustering, calculate the spatial distance and normal vector direction between each deviation point. When the normal vector direction is consistent and the spatial distance conforms to the logic of the wall length and width, the corresponding deviation points are regarded as the same wall.

6. The method for generating a three-dimensional floor plan model from a house surveying floor plan according to claim 4, characterized in that, The implementation of step S23 also includes: The difference planes are tracked, and when the current difference plane is exhausted, the shape of the difference plane is checked to determine the correctness of the shape of the difference plane during the exhaustive search. If the shape of the current difference plane is incorrectly identified, the exhaustive order during the exhaustive process of the difference plane is analyzed, and the structural overlap between the corresponding wall and the standard model is determined. The shape of the standard model with the maximum structural overlap is taken as the shape of the current difference plane. When the shape of the current difference plane is correctly identified, the corresponding difference plane is output according to the ratio of the tilt angle and center offset of the difference plane.

7. The method for generating a three-dimensional floor plan model from a house surveying floor plan according to claim 6, characterized in that, The implementation methods of step S31 include: The differences in the current input are analyzed, and based on the offset and tilt angle of the differences, each wall is classified into different processing methods. Determine if the current wall belongs to any one of the following categories: offset, tilt, or deformation. Generate the corresponding processing method and, based on the processing method corresponding to the current wall, obtain the deviation probability value for each wall in the random change matrix.

8. The method for generating a three-dimensional floor plan model from a house surveying floor plan according to claim 7, characterized in that, When collecting constraint conditions in step S32, the implementation method also includes: The constraints for each wall are recorded in the manner of wall-deviation dimension. The constraints are then filtered using the deviation probability and regarded as reference constraints.

9. The method for generating a three-dimensional floor plan model from a house surveying floor plan according to claim 1, characterized in that, The implementation of step S4 also includes: Retrieve the inflection points of the wall and the midpoints on the wall plane as the current set of coordinate points; align multiple sets of points in the coordinate set according to their included angles and verify their coordinates; after any verification condition is met, use the set of coordinate points corresponding to the verification condition as the current output data.

Images