A primitive identification method for building roof based on real noise inversion

Abstract

The application discloses a building roof Primitive identification method based on real noise inversion, relates to the field of real scene three-dimensional reconstruction, and solves the problem of poor identification effect of the existing building roof Primitive identification method, and comprises the following steps: S1, laser point cloud collection is performed on a target building roof to obtain a preselected roof point cloud set, roof laser sampling points in the preselected roof point cloud set are identified, and laser point cloud identification data are obtained; S2, the roof laser sampling points are analyzed in terms of aggregation according to the point cloud identification data, Primitive primitive identification is performed on the target building roof according to an analysis result, and regional primitive identification data are obtained; and S3, sampling point cross sections of roof Primitive primitives in the regional primitive identification data are collected to obtain a plurality of primitive sampling cross sections and perform interval analysis, the roof Primitive primitives are geometrically checked according to the analysis, and a checking result is fed back, and the application can improve the accuracy of building roof Primitive identification.

Description

Technical Field

[0001] This invention belongs to the field of real-scene 3D reconstruction and involves Primitive recognition technology, specifically a Primitive recognition method for building roofs based on real noise inversion. Background Technology

[0002] Existing methods for identifying building roof primitives have the following drawbacks when performing roof primitive identification: 1. Most existing roof primitive recognition methods based on synthetic point cloud training use idealized noise simulation methods such as Gaussian noise and fixed random perturbation to generate training samples, which are significantly different from the real noise morphology of oblique photogrammetric point clouds in real-world 3D engineering, resulting in a lack of accuracy in roof primitive recognition results. 2. Traditional roof primitive identification systems usually rely on finite standard roof templates to complete the fitting classification. When dealing with buildings with local small slopes, skylights, equipment bases, and composite roofs, they are prone to oversimplifying multi-component roofs into a single primitive, leading to misclassification. Even if a single surface is identified accurately, topological contradictions such as broken ridges, unclosed boundaries, and disordered slope height relationships often occur, making it difficult to guarantee the integrity of the overall roof structure.

[0003] To address this, we propose a method for identifying building rooftop primitives based on real noise inversion. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method for identifying building roof primitives based on real noise inversion, thereby improving the accuracy of building roof primitive identification methods.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for identifying building rooftop primitives based on real noise inversion, comprising the following steps: Step S1: Collect laser point clouds on the roof of the target building to obtain a set of pre-selected roof point clouds, and identify the roof laser sampling points in the set of pre-selected roof point clouds to obtain laser point cloud identification data; Step S2: Perform clustering analysis on the roof laser sampling points based on the point cloud recognition data, and perform Primitive primitive recognition on the target building roof based on the analysis results to obtain regional primitive recognition data; Step S3: Collect sampling point cross-sections of the roof primitive primitive in the regional primitive identification data, obtain multiple primitive sampling cross-sections and perform spacing analysis, perform geometric verification of the roof primitive primitive based on the analysis, and provide feedback on the verification results.

[0006] Furthermore, in step S1, the specific steps are as follows: Collect data on the rooftops of buildings that require Primitive identification to obtain the target building rooftops; The spatial area where the roof of the target building is located is collected to obtain a pre-selected spatial area, and a 3D model of the pre-selected spatial area is performed to obtain a 3D model of the pre-selected space. Using a lidar, a laser scan is performed on a pre-selected spatial area. Based on the scan results, laser sampling points of the spatial area to which the roof of the target building belongs are acquired to obtain a roof environmental point cloud set. The roof environmental point cloud set is then mapped onto the pre-selected spatial 3D model to obtain a supplementary 3D model of the spatial point cloud.

[0007] Furthermore, in step S1, the specific steps are as follows: In the spatial point cloud supplemented 3D model, a spatial coordinate system is created to obtain a pre-selected spatial coordinate system. In the spatial point cloud supplemented 3D model, the model space area covered by the building roof is discretized into several spatial pixels. The ordinate value of each spatial pixel in the pre-selected spatial coordinate system is obtained to obtain multiple spatial elevation values. The multiple spatial elevation values ​​are compared numerically, and the numerical interval formed by the minimum spatial elevation value and the maximum spatial elevation value is obtained to obtain the pre-selected roof elevation interval. Numerical acquisition of the ordinate of the laser sampling points in the rooftop environmental point cloud set in the pre-selected spatial coordinate system is performed to obtain the laser sampling elevation value corresponding to each laser sampling point. If the laser sampling elevation value is within the pre-selected rooftop elevation range, the corresponding laser sampling point is marked as a pre-selected area sampling point. If the laser sampling elevation value is not within the pre-selected rooftop elevation range, the corresponding laser sampling point is marked as a non-pre-selected area sampling point.

[0008] Furthermore, in step S1, the specific steps are as follows: The three-dimensional coordinates of the sampling points in the pre-selected area are collected in the pre-selected spatial coordinate system to obtain the three-dimensional coordinate values ​​of the sampling points. The spatial pixels within the building roof coverage area are collected to obtain multiple roof area spatial pixels. The three-dimensional coordinates of each roof area spatial pixel are collected in the pre-selected spatial coordinate system to obtain the roof area three-dimensional coordinate set. If the three-dimensional coordinates of the sampling point are within the three-dimensional coordinate set of the roof area, the corresponding pre-selected area sampling point is classified as a roof laser sampling point. If the three-dimensional coordinates of the sampling point are not within the three-dimensional coordinate set of the roof area, the corresponding pre-selected area sampling point is classified as a noise laser sampling point, thus obtaining laser point cloud recognition data.

[0009] Furthermore, in step S2, the specific steps are as follows: Step S21: Obtain laser point cloud recognition data. Based on the laser point cloud recognition data, obtain the roof laser sampling points contained in the roof of the target building to obtain multiple roof laser sampling points. Step S22: Randomly select a sample laser sampling point from the acquired roof laser sampling points, use the sample laser sampling point as the starting point of the primitive, perform sampling point clustering analysis, and obtain the Primitive primitive corresponding to the sample laser sampling point based on the analysis results; Step S23: Perform clustering analysis on the roof laser sampling points that have not been expanded to Primitive primitives. Based on the analysis, divide the roof laser sampling points into multiple Primitive primitives to obtain regional primitive identification data.

[0010] Furthermore, in step S22, the specific steps are as follows: In the spatial point cloud supplemented 3D model, the distance value between each roof laser sampling point and the sample laser sampling point is obtained. The roof laser sampling points are marked as T1 neighboring laser sampling points to Ta neighboring laser sampling points according to the descending order of the distance value. The line determined by the sample laser sampling point and the adjacent laser sampling point T1 is collected to obtain the initial line of the sampling point. If the adjacent laser sampling point T2 does not intersect with the initial line of the sampling point, the spatial plane determined by the sample laser sampling point, the adjacent laser sampling point T1, and the adjacent laser sampling point T2 is marked as the initial sampling plane of the roof. If the adjacent laser sampling point T2 intersects with the initial line of the sampling point, the adjacent laser sampling point T3 is used to replace the adjacent laser sampling point T2. This process continues until the replaced adjacent laser sampling point does not intersect with the initial line of the sampling point.

[0011] Furthermore, in step S22, the specific steps are as follows: The roof laser sampling points not covered by the initial roof sampling plane are collected. If the roof laser sampling points are coplanar with the initial roof sampling plane, the roof laser sampling points are directly used to expand the coplanarity of the initial roof sampling plane. If the roof laser sampling point is not coplanar with the initial roof sampling plane, then obtain the distance value between the roof laser sampling point and the initial roof sampling plane, and mark the roof laser sampling point as Q1 feature sampling point to Qb feature sampling point in ascending order of distance value; The initial sampling plane of the roof is discretized into several edge sampling points. The distance between each edge sampling point and the Q1 feature sampling point is obtained. The obtained distance values ​​are compared, and the edge sampling point corresponding to the minimum distance is marked as the near edge sampling point. The near edge sampling point is connected with the Q1 feature sampling point to obtain the feature sampling line. The angle value of the angle between the feature sampling line and the initial sampling plane of the roof at the near edge sampling point is collected to obtain the point-plane angle value corresponding to the feature laser sampling point. A preset range of point-plane angle values ​​is set. If the point-plane angle value is within the preset range, the Q1 feature sampling point is expanded to the initial sampling plane of the roof.

[0012] Furthermore, in step S22, the specific steps are as follows: If the angle between the point and the plane is not within the preset range of the angle between the point and the plane, the Q1 feature sampling point is removed and replaced by the Q2 feature sampling point until the expansion of the initial sampling plane of the roof is completed, and the Primitive primitive corresponding to the sample laser sampling point is obtained. If all feature sampling points from Q2 to Qb have completed the expansion of feature sampling point Q1, and no new feature sampling points have been added to the initial roof sampling plane, then the original initial roof sampling plane is set as the Primitive primitive corresponding to the sample laser sampling point; If the Q1 feature sampling point is expanded to the initial roof sampling plane, the expanded sampling plane is used to replace the initial roof sampling plane. This process continues until the expansion judgment of the Qb feature sampling point is completed. The final expanded graphic is then set as the Primitive primitive corresponding to the sample laser sampling point.

[0013] Furthermore, in step S3, the specific steps are as follows: Acquire region primitive recognition data, collect primitive primitives based on the region primitive recognition data, and arbitrarily select a sample primitive from the acquired primitive primitives; The expanded sampling points corresponding to the sample primitives are collected to obtain the initial sampling plane of the roof. A plane parallel to the initial sampling plane of the roof is drawn through each expanded sampling point to obtain multiple sampling parallel planes. Planar distances are collected for adjacent parallel sampling planes to obtain sampling plane distance values. The variance of the obtained sampling plane distance values ​​is then calculated to obtain the sampling plane distance variance.

[0014] Furthermore, in step S3, the specific steps are as follows: Create a preset value for the variance of the sampling surface distance. If the variance of the sampling surface distance is greater than the preset value, mark the sample primitive on the noise roof primitive and provide noise feedback. If the variance of the sampling surface distance is less than or equal to the preset value of the variance of the sampling surface distance, the sample element will be marked on the homogeneous roof element. Perform geometric verification on each roof primitive element and provide feedback on the verification results.

[0015] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. This invention first completes the identification of roof laser sampling points, then conducts sampling point aggregation analysis and performs Primitive primitive identification. Through a hierarchical and progressive processing logic, it gradually filters the effective roof point set, which can filter out the interference of discrete noise points such as vegetation and roof equipment in advance. Relying on the aggregation characteristics of point cloud, it distinguishes between continuous roof areas and scattered interference points, which can reduce the negative impact of noise data on primitive identification, reduce problems such as single patch adhesion and primitive missegmentation, improve the accuracy of roof Primitive primitive division, and adapt to complex roof laser point cloud scenarios with protruding components and occlusion interference. 2. This invention extracts sampling sections from identified roof primitives and performs section spacing analysis. It then uses the dispersion of section spacing to perform geometric verification of the primitives, along with a corresponding verification result feedback mechanism. Compared to the traditional method of judging patch quality solely based on the residual of a single plane fitting, section spacing quantifies the dispersion of sampling points within the primitive along the roof normal, accurately identifying abnormal primitives with mixed noise and uneven surfaces. Attached Figure Description

[0016] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0017] Figure 1 This is a schematic diagram of the pre-selected roof elevation range of the present invention; Figure 2 This is a schematic diagram of the sampling parallel plane of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1 This invention provides a technical solution: a method for identifying building rooftop primitives based on real noise inversion, comprising the following steps: Step S1: Collect laser point clouds on the roof of the target building to obtain a set of pre-selected roof point clouds, and identify the roof laser sampling points in the set of pre-selected roof point clouds to obtain laser point cloud identification data; In step S1, the specific steps are as follows: Collect data on the rooftops of buildings that require Primitive identification to obtain the target building rooftops; It should be noted here that: In this application, the Primitive referred to herein is a system-preset, parameter-controllable, and indivisible standard basic geometric unit.

[0020] The spatial area where the roof of the target building is located is collected to obtain a pre-selected spatial area, and a 3D model of the pre-selected spatial area is performed to obtain a 3D model of the pre-selected space. Using a lidar, a laser scan is performed on a pre-selected spatial area. Based on the scan results, laser sampling points of the spatial area to which the roof of the target building belongs are acquired to obtain a roof environmental point cloud set. The roof environmental point cloud set is then mapped onto a pre-selected spatial 3D model to obtain a supplementary 3D model of the spatial point cloud. Please see Figure 1 In the spatial point cloud supplemented 3D model, a spatial coordinate system is created to obtain a pre-selected spatial coordinate system. In the spatial point cloud supplemented 3D model, the model space area covered by the building roof is discretized into several spatial pixels. The ordinate value of each spatial pixel in the pre-selected spatial coordinate system is obtained to obtain multiple spatial elevation values. The multiple spatial elevation values ​​are compared to obtain the numerical interval formed by the minimum spatial elevation value and the maximum spatial elevation value, thus obtaining the pre-selected roof elevation interval.

[0021] Numerical acquisition of the ordinate of the laser sampling points in the rooftop environmental point cloud set in the pre-selected spatial coordinate system is performed to obtain the laser sampling elevation value corresponding to each laser sampling point. If the laser sampling elevation value is within the pre-selected rooftop elevation range, the corresponding laser sampling point is marked as a pre-selected area sampling point. If the laser sampling elevation value is not within the pre-selected rooftop elevation range, the corresponding laser sampling point is marked as a non-pre-selected area sampling point. It should be noted here that: In this application, the pre-selected area sampling points referred to herein are specifically laser sampling points whose elevation values ​​fall within the corresponding elevation range of the roof area.

[0022] The three-dimensional coordinates of the sampling points in the pre-selected area are collected in the pre-selected spatial coordinate system to obtain the three-dimensional coordinate values ​​of the sampling points. The spatial pixels within the building roof coverage area are collected to obtain multiple roof area spatial pixels. The three-dimensional coordinates of each roof area spatial pixel are collected in the pre-selected spatial coordinate system to obtain the roof area three-dimensional coordinate set. If the three-dimensional coordinates of the sampling point are within the three-dimensional coordinate set of the roof area, the corresponding pre-selected area sampling point is classified as a roof laser sampling point. If the three-dimensional coordinates of the sampling point are not within the three-dimensional coordinate set of the roof area, the corresponding pre-selected area sampling point is classified as a noise laser sampling point, thus obtaining laser point cloud recognition data. Step S2: Perform clustering analysis on the roof laser sampling points based on the point cloud recognition data, and perform Primitive primitive recognition on the target building roof based on the analysis results to obtain regional primitive recognition data; In step S2, the specific steps are as follows: Acquire laser point cloud recognition data, and obtain multiple roof laser sampling points based on the laser point cloud recognition data. Arbitrarily select one sample laser sampling point from the acquired roof laser sampling points, and use the sample laser sampling point as the starting point of the primitive for sampling point clustering analysis. Based on the analysis results, obtain the Primitive primitive corresponding to the sample laser sampling point.

[0023] Specifically as follows: In the spatial point cloud supplemented 3D model, the distance value between each roof laser sampling point and the sample laser sampling point is obtained. The roof laser sampling points are marked as T1 neighboring laser sampling points to Ta neighboring laser sampling points according to the descending order of the distance value. It should be noted here that: In this application, T1 to Ta in T1 to Ta are the marking symbols corresponding to the neighboring laser sampling points; In this application, if the roof laser sampling point and the sample laser sampling point are in an intersecting state, and the distance between the roof laser sampling point and the sample laser sampling point is 0.

[0024] The line determined by the sample laser sampling point and the adjacent laser sampling point T1 is collected to obtain the initial line of the sampling point. If the adjacent laser sampling point T2 does not intersect with the initial line of the sampling point, the spatial plane determined by the sample laser sampling point, the adjacent laser sampling point T1, and the adjacent laser sampling point T2 is marked as the initial sampling plane of the roof. If the adjacent laser sampling point T2 intersects with the initial line of the sampling point, the adjacent laser sampling point T3 is used to replace the adjacent laser sampling point T2. This process continues until the replaced adjacent laser sampling point does not intersect with the initial line of the sampling point.

[0025] The roof laser sampling points not covered by the initial roof sampling plane are collected. If the roof laser sampling points are coplanar with the initial roof sampling plane, the roof laser sampling points are directly used to expand the coplanarity of the initial roof sampling plane. If the roof laser sampling point is not coplanar with the initial roof sampling plane, then obtain the distance value between the roof laser sampling point and the initial roof sampling plane, and mark the roof laser sampling point as Q1 feature sampling point to Qb feature sampling point in ascending order of distance value.

[0026] It should be noted here that: In this application, Q1 to Qb in the feature sampling points Q1 to Qb are the corresponding marker symbols of the feature sampling points; In this application, the distances between the Q1 feature sampling point and the Qb feature sampling point and the initial sampling plane of the roof are all less than the preset feature point surface distance, which is 5mm.

[0027] The initial sampling plane of the roof is discretized into several edge sampling points. The distance between each edge sampling point and the Q1 feature sampling point is obtained. The obtained distance values ​​are compared, and the edge sampling point corresponding to the minimum distance is marked as the near edge sampling point. The near edge sampling point is connected with the Q1 feature sampling point to obtain the feature sampling line. The angle value of the angle between the feature sampling line and the initial sampling plane of the roof at the near edge sampling point is collected to obtain the point-plane angle value corresponding to the feature laser sampling point. A preset range of point-plane angle values ​​is set. If the point-plane angle value is within the preset range, the Q1 feature sampling point is expanded to the initial sampling plane of the roof.

[0028] It should be noted here that: In this application, laser sampling points are acquired from the historical expansion to the initial sampling plane of the roof. The point-to-plane angle value corresponding to each laser sampling point is acquired. The acquired point-to-plane angle values ​​are compared. The maximum point-to-plane angle value is marked as the upper limit of the preset range of point-to-plane angle values, and the minimum point-to-plane angle value is marked as the lower limit of the preset range of point-to-plane angle values.

[0029] If the angle between the point and the plane is not within the preset range of the angle between the point and the plane, the Q1 feature sampling point is removed and replaced by the Q2 feature sampling point until the expansion of the initial sampling plane of the roof is completed, and the Primitive primitive corresponding to the sample laser sampling point is obtained. If all feature sampling points from Q2 to Qb have completed the expansion of feature sampling point Q1, and no new feature sampling points have been added to the initial roof sampling plane, then the original initial roof sampling plane is set as the Primitive primitive corresponding to the sample laser sampling point; If the Q1 feature sampling point is expanded to the initial roof sampling plane, the expanded sampling plane is used to replace the initial roof sampling plane. This process continues until the expansion judgment of the Qb feature sampling point is completed. The final expanded graphic is then set as the Primitive primitive corresponding to the sample laser sampling point.

[0030] Clustering analysis was performed on the roof laser sampling points that were not expanded to Primitive primitives. Based on the analysis, the roof laser sampling points were segmented into multiple Primitive primitives to obtain regional primitive identification data.

[0031] It should be noted here that: In this application, the Primitive primitives referred to herein are individual topological patches segmented from the roof of the target building. The Primitive primitives in this application no longer refer to general basic geometric shapes such as cubes and cones, but rather to individual roof topological patches that are independent of each other, have clear boundaries, and are continuous and flat after the point cloud of the entire building roof is segmented according to the slope characteristics. For example, each independent slope is a roof Primitive primitive.

[0032] Step S3: Collect sampling point cross-sections of the roof primitive primitive in the regional primitive identification data, obtain multiple primitive sampling cross-sections and perform spacing analysis, perform geometric verification of the roof primitive primitive based on the analysis, and provide feedback on the verification results.

[0033] In step S3, the specific steps are as follows: Acquire region primitive recognition data, collect primitive primitives based on the region primitive recognition data, and arbitrarily select a sample primitive from the acquired primitive primitives; The expanded sampling points corresponding to the sample primitives are collected to obtain the initial sampling plane of the roof. A plane parallel to the initial sampling plane of the roof is drawn through each expanded sampling point to obtain multiple sampling parallel planes. Planar distances are collected for adjacent parallel sampling planes to obtain sampling plane distance values. The variance of the obtained sampling plane distance values ​​is then calculated to obtain the sampling plane distance variance. It should be noted here that: Specifically, adjacent sampling parallel planes are defined as follows: all sampling parallel planes are sorted along the roof normal direction according to their vertical distance from the reference plane in ascending order. Two planes in the sorted sequence that have no other sampling parallel planes in between are considered adjacent sampling parallel planes. (See also...) Figure 2 The sampling parallel planes W1 and W2 are adjacent sampling parallel planes. The sampling parallel plane W2 is adjacent to both the sampling parallel planes W1 and W2. The sampling parallel plane W3 is adjacent to both the sampling parallel planes W2 and W4.

[0034] Create a preset value for the variance of the sampling surface. If the variance of the sampling surface is greater than the preset value, mark the sample element on the noisy roof element and provide noise feedback. If the variance of the sampling surface is less than or equal to the preset value, mark the sample element on the homogeneous roof element. It should be noted here that: In this application, the sampling surface distance variance is calculated for historical Primitive elements that have been identified as homogeneous roof elements, and the obtained sampling surface distance variance is numerically compared. The maximum sampling surface distance variance is marked as the preset value of sampling surface distance variance. In this application, the initial sampling plane of the roof is fitted by expanding the sampling points of the sample primitive as a reference. A series of parallel planes are constructed through each sampling point. The vertical distance between adjacent parallel planes can characterize the offset of each sampling point relative to the reference plane along the roof normal. The variance of the sampling distance can quantify the degree of dispersion of the sampling points within the primitive along the roof normal direction. If the variance of the sampling distance is greater than the preset value, it indicates that the sampling points inside the primitive are randomly distributed, and mostly contain roof protrusions, vegetation spots, or multi-slope adhered surfaces, and do not have the characteristics of a uniform and flat roof. Therefore, it is marked as a noisy roof primitive and noise feedback is output for subsequent removal processing. If the variance of the sampling distance is less than or equal to the preset value, it means that the sampling points are uniformly offset along the normal direction and the local slope is flat and uniform, which meets the geometric characteristics of a standard roof. Therefore, it is marked as a homogeneous roof primitive. Perform geometric verification on each roof primitive element and provide feedback on the verification results.

[0035] Compared to the problems described in the background technology, this invention first completes the identification of roof laser sampling points, then performs sampling point aggregation analysis and performs primitive primitive identification. Through a hierarchical and progressive processing logic, it gradually filters the effective roof point set, which can filter out the interference of discrete noise points such as vegetation and roof equipment in advance. Relying on the point cloud aggregation characteristics, it distinguishes between continuous roof areas and scattered interference points, which can reduce the negative impact of noise data on primitive identification, reduce problems such as single patch adhesion and primitive missegmentation, improve the accuracy of roof primitive primitive division, and adapt to complex roof laser point cloud scenarios with protruding components and occlusion interference. Furthermore, this invention extracts sampling sections from the identified roof primitive primitives and performs section spacing analysis. It performs geometric verification of the primitives based on the dispersion of section spacing, and also includes a verification result feedback mechanism.

[0036] Compared to the traditional method of judging the quality of a surface patch by relying solely on the residual of fitting a single plane, the cross-sectional spacing can quantify the degree of dispersion of the sampling points inside the primitive along the roof normal, and accurately identify abnormal primitives with mixed noise and unevenness.

[0037] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to specific implementations. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for identifying building rooftop primitives based on real noise inversion, characterized in that, include: Step S1: Collect laser point clouds on the roof of the target building to obtain a set of pre-selected roof point clouds, and identify the roof laser sampling points in the set of pre-selected roof point clouds to obtain laser point cloud identification data; Step S2: Perform clustering analysis on the roof laser sampling points based on the point cloud recognition data, and perform Primitive primitive recognition on the target building roof to obtain regional primitive recognition data; Step S3: Collect sampling point cross-sections of the roof primitive primitive in the regional primitive identification data, obtain multiple primitive sampling cross-sections and perform spacing analysis, perform geometric verification of the roof primitive primitive based on the analysis, and provide feedback on the verification results.

2. The method for identifying building rooftop primitives based on real noise inversion according to claim 1, characterized in that, In step S1, the specific steps are as follows: Collect data on the rooftops of buildings that require Primitive identification to obtain the target building rooftops; The spatial area where the roof of the target building is located is collected to obtain a pre-selected spatial area, and a 3D model of the pre-selected spatial area is performed to obtain a 3D model of the pre-selected space. Using a lidar, a laser scan is performed on a pre-selected spatial area to acquire laser sampling points in the spatial area where the target building's roof belongs, resulting in a roof environment point cloud set. This roof environment point cloud set is then mapped onto a pre-selected spatial 3D model to obtain a supplementary 3D model of the spatial point cloud.

3. The method for identifying building rooftop primitives based on real noise inversion according to claim 2, characterized in that, In step S1, the specific steps are as follows: In the spatial point cloud supplemented 3D model, a spatial coordinate system is created to obtain a pre-selected spatial coordinate system. In the spatial point cloud supplemented 3D model, the model space area covered by the building roof is discretized into several spatial pixels. The ordinate value of each spatial pixel in the pre-selected spatial coordinate system is obtained to obtain multiple spatial elevation values. The numerical interval formed by the minimum spatial elevation value and the maximum spatial elevation value is collected to obtain the pre-selected roof elevation interval. Numerical acquisition of the ordinate of the laser sampling points in the rooftop environmental point cloud set in the pre-selected spatial coordinate system is performed to obtain the laser sampling elevation value corresponding to each laser sampling point. If the laser sampling elevation value is within the pre-selected rooftop elevation range, the corresponding laser sampling point is marked as a pre-selected area sampling point. If the laser sampling elevation value is not within the pre-selected rooftop elevation range, the corresponding laser sampling point is marked as a non-pre-selected area sampling point.

4. The method for identifying building rooftop primitives based on real noise inversion according to claim 3, characterized in that, In step S1, the specific steps are as follows: The three-dimensional coordinates of the sampling points in the pre-selected area are collected in the pre-selected spatial coordinate system to obtain the three-dimensional coordinate values ​​of the sampling points. The spatial pixels within the building roof coverage area are collected to obtain multiple roof area spatial pixels. The three-dimensional coordinates of each roof area spatial pixel are collected in the pre-selected spatial coordinate system to obtain the roof area three-dimensional coordinate set. If the three-dimensional coordinates of the sampling point are within the three-dimensional coordinate set of the roof area, the corresponding pre-selected area sampling point is classified as a roof laser sampling point. If the three-dimensional coordinates of the sampling point are not within the three-dimensional coordinate set of the roof area, the corresponding pre-selected area sampling point is classified as a noise laser sampling point, thus obtaining laser point cloud recognition data.

5. The method for identifying building rooftop primitives based on real noise inversion according to claim 1, characterized in that, In step S2, the specific steps are as follows: Step S21: Obtain laser point cloud recognition data. Based on the laser point cloud recognition data, obtain the roof laser sampling points contained in the roof of the target building to obtain multiple roof laser sampling points. Step S22: Randomly select a sample laser sampling point from the obtained roof laser sampling points, and use the sample laser sampling point as the starting point of the primitive to perform sampling point clustering analysis to obtain the Primitive primitive corresponding to the sample laser sampling point; Step S23: Perform clustering analysis on the roof laser sampling points that have not been expanded to Primitive primitives. Based on the analysis, divide the roof laser sampling points into multiple Primitive primitives to obtain regional primitive identification data.

6. The method for identifying building rooftop primitives based on real noise inversion according to claim 5, characterized in that, In step S22, the specific steps are as follows: In the spatial point cloud supplemented 3D model, the distance value between each roof laser sampling point and the sample laser sampling point is obtained. The roof laser sampling points are marked as T1 neighboring laser sampling points to Ta neighboring laser sampling points according to the descending order of the distance value. The line determined by the sample laser sampling point and the adjacent laser sampling point T1 is collected to obtain the initial line of the sampling point. If the adjacent laser sampling point T2 does not intersect with the initial line of the sampling point, the spatial plane determined by the sample laser sampling point, the adjacent laser sampling point T1, and the adjacent laser sampling point T2 is marked as the initial sampling plane of the roof. If the adjacent laser sampling point T2 intersects with the initial line of the sampling point, the adjacent laser sampling point T3 is used to replace the adjacent laser sampling point T2 until the replaced adjacent laser sampling point does not intersect with the initial line of the sampling point.

7. The method for identifying building rooftop primitives based on real noise inversion according to claim 6, characterized in that, In step S22, the specific steps are as follows: The roof laser sampling points not covered by the initial roof sampling plane are collected. If the roof laser sampling points are coplanar with the initial roof sampling plane, the roof laser sampling points are directly used to expand the coplanarity of the initial roof sampling plane. If the roof laser sampling point is not coplanar with the initial roof sampling plane, then obtain the distance value between the roof laser sampling point and the initial roof sampling plane, and mark the roof laser sampling point as Q1 feature sampling point to Qb feature sampling point in ascending order of distance value; The initial sampling plane of the roof is discretized into several edge sampling points. The distance between each edge sampling point and the Q1 feature sampling point is obtained. The edge sampling point corresponding to the minimum distance is marked as the near edge sampling point. The near edge sampling point is connected to the Q1 feature sampling point to obtain the feature sampling line. The angle value of the angle between the feature sampling line and the initial sampling plane of the roof at the near edge sampling point is collected to obtain the point-to-surface angle value corresponding to the feature laser sampling point. A preset range of point-to-surface angle values ​​is set. If the point-to-surface angle value is within the preset range, the Q1 feature sampling point is expanded to the initial sampling plane of the roof.

8. The method for identifying building rooftop primitives based on real noise inversion according to claim 7, characterized in that, In step S22, the specific steps are as follows: If the angle between the point and the plane is not within the preset range of the angle between the point and the plane, the Q1 feature sampling point is removed and replaced by the Q2 feature sampling point until the expansion of the initial sampling plane of the roof is completed, and the Primitive primitive corresponding to the sample laser sampling point is obtained. If the Q1 feature sampling point is expanded to the initial roof sampling plane, the expanded sampling plane is used to replace the initial roof sampling plane until the expansion judgment of the Qb feature sampling point is completed, and the final expanded graphic is set as the Primitive primitive corresponding to the sample laser sampling point.

9. The method for identifying building rooftop primitives based on real noise inversion according to claim 1, characterized in that, In step S3, the specific steps are as follows: Acquire region primitive recognition data, collect primitive primitives based on the region primitive recognition data, and arbitrarily select a sample primitive from the acquired primitive primitives; The expanded sampling points corresponding to the sample primitives are collected to obtain the initial sampling plane of the roof. A plane parallel to the initial sampling plane of the roof is drawn through each expanded sampling point to obtain multiple sampling parallel planes. Planar distances are collected for adjacent parallel sampling planes to obtain sampling plane distance values. The variance of the obtained sampling plane distance values ​​is then calculated to obtain the sampling plane distance variance.

10. The method for identifying building rooftop primitives based on real noise inversion according to claim 9, characterized in that, In step S3, the specific steps are as follows: Create a preset value for the variance of the sampling surface distance. If the variance of the sampling surface distance is greater than the preset value, mark the sample primitive on the noise roof primitive and provide noise feedback. If the variance of the sampling surface distance is less than or equal to the preset value of the variance of the sampling surface distance, the sample element will be marked on the homogeneous roof element. Perform geometric verification on each roof primitive element and provide feedback on the verification results.

Images