A method and system for extracting building roof surface point cloud from airborne lidar data

Abstract

This application proposes a method and system for extracting building roof surface point clouds from airborne laser radar data, which relates to the field of airborne laser radar technology, including: using a cloth simulation filtering algorithm to distinguish ground points from non-ground points; using a seed point extraction algorithm based on neighborhood information to extract a seed point set with roof surface semantic information; classifying and growing vegetation points and non-vegetation points under different growth constraint conditions to preliminarily obtain roof surface point clouds; using the height and area values ​​of the roof surface to filter and optimize the roof surface point cloud extraction results, and finally obtaining accurate building roof surface point cloud data.

Description

Technical Field

[0001] The present application relates to the field of airborne laser radar technology, and more specifically, to a method and system for extracting building roof surface point clouds from airborne laser radar data. Background Art

[0002] Airborne LiDAR (Light Detection and Ranging) technology efficiently captures high-density 3D point cloud data of roads, buildings, vegetation, and other man-made structures, providing crucial data support for current digital urban development. Accurately segmenting roof point clouds is crucial for reconstructing 3D building models.

[0003] Methods for extracting roof point clouds from airborne LiDAR data can be broadly categorized into three categories: First, feature clustering-based methods. Zhao et al. used principal component analysis to extract roof seed points and then clustered the local normal vector distribution density features to extract roof surfaces. Sun et al. used graph segmentation to remove vegetation areas and then used hierarchical Euclidean clustering to extract building roof patches. However, these methods still suffer from difficulties in neighborhood selection and sensitivity to noise and outliers, as neighborhood selection significantly impacts the quality of point attribute calculations. Second, model fitting-based methods. YI et al. combined a random sampling consensus algorithm with a Bayesian probability model to extract building roof point clouds. Kang Chuanli et al. proposed a roof extraction method combining SIFT and 3D Hough transform, using 3D Hough space to determine the reliability of fitted planes and thereby reduce the occurrence of spurious planes. While this method is highly stable, it is prone to generating a high number of spurious planes in locally flat areas of vegetation canopy, and extraction efficiency decreases significantly with increasing data complexity. Third, region growing-based methods are used. SHAO et al. used cloth simulation and roughness features to detect seed points, and then used the normal vector angle as a growth constraint to perform growth segmentation of the roof surface. KAN et al. used curvature features to extract roof seed points in a hierarchical manner, and introduced distance constraints on the basis of traditional normal vector angle constraints to reduce the misclassification of lush vegetation points. He Manyun et al. proposed a regional growth method based on irregular triangulated networks. First, the roof edge points were extracted using angle mesh features, and then growth segmentation was performed based on the connection relationship between angle meshes to obtain the roof surface point cloud. This type of method is simple to apply and has high computational efficiency, but it has high requirements for the extraction of seed points, and is prone to severe under-segmentation when there is adhesion or occlusion between buildings and vegetation.

[0004] In summary, the existing building roof point cloud extraction algorithm based on airborne LiDAR data has low extraction accuracy and is easily affected by noise, especially when the vegetation is lush, which easily leads to the misclassification of vegetation points. Summary of the Invention

[0005] The purpose of this application is to provide a building roof surface point cloud extraction method from airborne lidar data, which can effectively distinguish roof points from vegetation points by introducing vegetation index and RGB color difference constraints, thereby ensuring the accuracy of roof surface growth segmentation.

[0006] Another object of the present application is to provide a building roof surface point cloud extraction system based on airborne laser radar data, which can run a building roof surface point cloud extraction method based on airborne laser radar data.

[0007] The embodiment of the present application is implemented as follows:

[0008] In the first aspect, an embodiment of the present application provides a method for extracting building roof surface point clouds from airborne lidar data, which includes using a cloth simulation filtering algorithm to distinguish ground points from non-ground points; using a seed point extraction algorithm based on neighborhood information to extract a seed point set with roof surface semantic information; performing classification growth segmentation on vegetation points and non-vegetation points under different growth constraint conditions to preliminarily obtain a roof surface point cloud; using the height and area values ​​of the roof surface to filter and optimize the roof surface point cloud extraction results, and finally obtaining accurate building roof surface point cloud data.

[0009] In some embodiments of the present application, the above-mentioned seed point extraction algorithm based on neighborhood information is used to extract a seed point set with roof surface semantic information, including: extracting constraints through neighborhood curvature statistics, neighborhood maximum distance and neighborhood gradient analysis based on the neighborhood characteristics of the roof surface point cloud.

[0010] In some embodiments of the present application, the above-mentioned neighborhood curvature statistics include: obtaining the normal vector and curvature by calculating the eigenvalues ​​and eigenvectors of the neighborhood covariance matrix, and calculating the point cloud curvature value; traversing all curvature values, retaining the curvature threshold less than the preset point, obtaining a set of candidate seed points, and then performing a neighbor search on the candidate seed points to achieve plane point extraction.

[0011] In some embodiments of the present application, the aforementioned neighborhood maximum distance includes: eliminating, by using the neighborhood maximum distance constraint, interfering planes with narrow and long features existing in the planar point cloud extracted using the neighborhood curvature statistical constraint.

[0012] In some embodiments of the present application, the neighborhood gradient analysis includes: distinguishing between the roof surface and the building facade based on different characteristics of the roof surface and the building facade using neighborhood gradient information.

[0013] In some embodiments of the present application, the classification growth segmentation of vegetation points and non-vegetation points under different growth constraint conditions is performed to initially obtain the roof surface point cloud, including: setting a loose normal vector angle threshold Ensure the integrity of building roof extraction, and then use the combined vegetation index and RGB color difference As an additional constraint, vegetation points and non-vegetation points are classified to solve the under-segmentation problem caused by point cloud segmentation.

[0014] In some embodiments of the present application, the above also includes: after the region growing segmentation is completed, the segmentation results are filtered and optimized using the height and area values ​​of the roof surface.

[0015] In a second aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method such as any one of the methods for extracting a building roof surface point cloud from airborne lidar data.

[0016] In a third aspect, an embodiment of the present application provides a building roof point cloud extraction system based on airborne lidar data, characterized by comprising:

[0017] Data preprocessing module, used to distinguish ground points from non-ground points using cloth simulation filtering algorithm;

[0018] The roof point cloud extraction module uses a seed point extraction algorithm based on neighborhood information to extract a set of seed points with semantic information about the roof surface. It then performs classification and growth segmentation on vegetation points and non-vegetation points under different growth constraints to initially obtain a roof point cloud.

[0019] The data post-processing module performs classification, growth and segmentation to initially obtain the roof surface point cloud; finally, the roof surface height and area values ​​are used to filter and optimize the roof surface point cloud extraction results, and ultimately obtain accurate building roof surface point cloud data.

[0020] Compared with the prior art, the embodiments of the present application have at least the following advantages or beneficial effects:

[0021] This algorithm effectively extracts robust rooftop seed points, avoiding both false and missed points. The segmented rooftops are complete and clearly defined, avoiding both under- and over-segmentation. The proposed algorithm is simple in principle and computationally efficient. It effectively addresses the problem of low extraction accuracy caused by misclassification of vegetation points. The algorithm achieves excellent rooftop point cloud extraction and is adaptable to diverse architectural scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 A schematic diagram of the steps of a method for extracting building roof surface point clouds from airborne LiDAR data provided in an embodiment of the present application;

[0024] Figure 2 A schematic diagram of the neighborhood curvature distribution of a plane point cloud and a vegetation point cloud provided in an embodiment of the present application;

[0025] Figure 3 A schematic diagram showing a comparison of the maximum distances between neighbors of different planar objects provided in an embodiment of the present application;

[0026] Figure 4 A schematic diagram of gradient value calculation provided in an embodiment of the present application;

[0027] Figure 5 A schematic diagram of over-segmentation and under-segmentation provided in an embodiment of the present application;

[0028] Figure 6 A schematic diagram of a neighborhood expansion process provided in an embodiment of the present application;

[0029] Figure 7 A schematic diagram of original point cloud data of a test area provided in an embodiment of the present application;

[0030] Figure 8 A schematic diagram of roof surface point cloud reference data provided in an embodiment of the present application;

[0031] Figure 9 A schematic diagram of non-ground point cloud data provided in an embodiment of the present application;

[0032] Figure 10 A schematic diagram of roof seed point data provided in an embodiment of the present application;

[0033] Figure 11 A schematic diagram of a roof surface point cloud extraction result provided in an embodiment of the present application;

[0034] Figure 12 A schematic diagram of an extraction result and a reference point cloud overlay comparison provided in an embodiment of the present application;

[0035] Figure 13 An electronic device provided in an embodiment of the present application;

[0036] Figure 14 A schematic diagram of a building roof surface point cloud extraction system module based on airborne lidar data provided in an embodiment of the present application.

[0037] Icon: 10-data preprocessing module; 20-roof surface point cloud extraction module; 30-data post-processing module; 101-memory; 102-processor; 103-communication interface. Implementation Method

[0038] To make the objectives, technical solutions, and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Generally, the components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations.

[0039] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for protection, but merely represents selected embodiments of the present application. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments in the present application without creative work are within the scope of protection of the present application.

[0040] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0041] It should be noted that the term "comprises" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or apparatus that includes a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0042] The following describes some embodiments of the present application in detail with reference to the accompanying drawings. In the absence of conflict, the following embodiments and features thereof may be combined with each other.

[0043] Example 1

[0044] See also Figure 1 , Figure 1 A schematic diagram of the steps of a method for extracting building roof surface point clouds from airborne LiDAR data provided in an embodiment of the present application is as follows:

[0045] Figure 1The overall process of the algorithm of this embodiment is demonstrated. The method is divided into three stages: data preprocessing, roof surface point cloud extraction, and data post-processing. Among them, the two steps of roof surface seed point extraction and classification region growth in stage two are the core content of this embodiment. The algorithm is described as follows: First, the cloth simulation filtering algorithm is used to distinguish ground points from non-ground points; secondly, the seed point extraction algorithm based on neighborhood information is used to extract the seed point set with roof surface semantic information; then, vegetation points and non-vegetation points are classified and grown under different growth constraints to preliminarily obtain the roof surface point cloud; finally, the roof surface height and area values ​​are used to filter and optimize the roof surface point cloud extraction results, and finally accurate building roof surface point cloud data is obtained.

[0046] The non-ground point cloud data obtained after filtering mainly includes building points, vegetation points, and other man-made facility points (vehicle points, wall points, power facility points, transportation facility points, etc.), among which buildings are composed of two parts: the building facade and the building roof. Building roofs are usually flat, so plane features are used to extract roof surface seed points. However, if only plane features are used as extraction criteria, it is easy to mistakenly classify vegetation canopy points and other man-made plane facility points with flat local surfaces as roof surface seed points, resulting in the final seed points being less robust. To address the above problems, this embodiment proposes a roof surface seed point extraction method using neighborhood information. Based on the neighborhood features of the roof surface point cloud, this method adopts a total of three extraction constraints, the details of which are as follows.

[0047] (1) Neighborhood curvature statistics

[0048] Since the distribution patterns of plane point clouds and vegetation point clouds are different, the curvature distributions are also different. Figure 2 (ab) are schematic diagrams of plane and vegetation point cloud data, respectively, where the green points are arbitrary points , and the red points are its neighboring points. Figure 2 (c) is the point The results of a statistical analysis of the curvature values ​​of plane points and their neighborhood points show that the curvature values ​​of plane points are much smaller than those of vegetation points, as shown in the boxplot. The normal curve shows that the curvature values ​​of the neighborhoods of plane points are relatively concentrated, while those of the neighborhoods of vegetation points are more dispersed. Based on this characteristic, vegetation points can be effectively eliminated while plane points are retained.

[0049] The specific implementation steps are as follows:

[0050] 1) Calculation of point cloud curvature value: Let the k-neighborhood point set of point p be The normal vector is obtained by calculating the eigenvalues ​​and eigenvectors of the neighborhood covariance matrix and curvature .

[0051] (1)

[0052] Where, is the eigenvalue of the covariance matrix; for The corresponding eigenvector.

[0053] 2) Plane point extraction: First, traverse all curvature values ​​and retain those with a curvature threshold less than Points, get the candidate seed point set Then, for the candidate seed points , , do nearest neighbor search, the neighborhood point set is The neighborhood points satisfy The number of conditions is n, if , then Add to seed point set Among them It is the plane coefficient, and its value range is generally 0.5~1.0. The larger the value, the higher the plane accuracy.

[0054] (2) Maximum neighborhood distance

[0055] In addition to rooftops, the planar point cloud extracted using neighborhood curvature statistical constraints also contains a large number of interfering planes, such as bus tops, wall tops, and neatly arranged greenery points. Analysis shows that compared to rooftops, these interfering planes are smaller and narrower. Figure 3 The maximum neighborhood distance comparison diagram of different plane objects, where the green point is the seed point to be judged and the red point is its neighborhood point, with the maximum neighborhood radius Represents the maximum distance of the neighborhood points. It can be seen that the maximum neighborhood radius of the roof surface point cloud (Figure a) It is obviously smaller than the maximum neighborhood radius of the narrow and long plane point cloud (Figures b and d). Based on this, this embodiment uses the maximum neighborhood distance constraint to eliminate the narrow and long interference planes.

[0056] The specific implementation steps are as follows:

[0057] 1) Euclidean distance calculation. Let the k-neighborhood point set of point p be .point To neighboring points distance Calculated using the Euclidean distance formula.

[0058] (2)

[0059] In the formula Neighborhood points Coordinate values, is the coordinate value of any point p. The distance of all neighboring points is calculated to obtain the distance set of neighboring points .

[0060] 2) Maximum distance constraint. Sort the elements in the neighborhood to get the maximum distance .like , then point p satisfies the seed point condition. It is the maximum neighborhood radius of the roof seed point, which is related to the number of neighborhood points and the density of point cloud data.

[0061] (3) Neighborhood gradient analysis

[0062] Airborne LiDAR data contains numerous building facade points. This example primarily focuses on rooftops, which serve as interfering surfaces and need to be removed. Based on the different characteristics of rooftops and building facades, this example uses neighborhood gradient information to distinguish between the two.

[0063] The specific implementation steps are as follows:

[0064] 1) Gradient value calculation: set point of The neighborhood point set is . Sorting the three coordinate components of the neighborhood point set in ascending order yields , , Three ordered sets, calculated , , , , , Six coordinate components.

[0065] (3)

[0066] use , The two coordinate points replace the neighborhood points to calculate the gradient value, and the three directional gradients of the neighborhood point set are obtained as follows: , , ,like Figure 4 shown.

[0067] 2) Neighborhood gradient constraint: If , It is an approximate replacement point of the neighborhood point set of the roof plane point, and its gradient value should satisfy the following relationship.

[0068] (4)

[0069] Where, is the Z-direction gradient coefficient. If all roof surfaces are flat, then Approaches 0, if the roof surface has an inclination angle, The value should increase as the roof pitch angle increases.

[0070] Using neighborhood curvature statistics to extract all planar points in the data, using neighborhood maximum distance constraints to remove narrow and long small interfering planar points, and using neighborhood gradient change constraints to remove building facade points. Points that meet all of the above requirements are defined as roof seed points.

[0071] Classification Region Growing:

[0072] Accurately segmenting rooftops requires not only robust seed points but also appropriate growth constraints. The normal vector angle is a fundamental constraint in traditional region growing. Using only this constraint for rooftop point cloud segmentation often results in over-segmentation or under-segmentation. Figure 5 (a) is the over-segmentation problem caused by setting the normal vector angle to a small value, which is manifested as incomplete roof surface extraction. Figure 5 (c) Under-segmentation occurs when the normal angle is set too high. This is manifested by the inclusion of some vegetation points in the segmented roof surface point cloud. This is particularly serious when buildings and vegetation are intertwined or blocked. Due to the complexity of point clouds in large scenes, it is difficult to achieve a good segmentation effect simply by adjusting the normal angle parameter.

[0073] Aiming at the above problems, a segmentation method based on classified region growing is proposed. Figure 6 This is the neighborhood expansion flowchart. This method first sets a loose normal vector angle threshold Ensure the integrity of building roof extraction; then use the combined vegetation index COM and RGB color difference As an additional constraint, vegetation points and non-vegetation points are classified to solve the under-segmentation problem caused by point cloud segmentation.

[0074] When the neighborhood of the seed point is expanded, the neighborhood points can be divided into two categories: one is the roof seed point itself, and the other is the natural vegetation point that is adhered to or blocked by the roof. These natural vegetation are mostly tall trees in the period of lush growth, and the vegetation surface has strong spectral reflectance characteristics in the visible light band

[13] . The roofs of artificial buildings are mainly composed of building materials such as concrete and bricks, and the colors are mostly red and gray. The surface spectral reflectance characteristics are not obvious. Therefore, the color index of the visible light band is used to reflect the difference between vegetation and buildings. This embodiment uses the combined vegetation index as the classification basis for vegetation points and roof points. The calculation formula is shown in Table 1 below:

[0075]

[0076] Where R, G, and B are the red, green, and blue channel values, respectively, and r, g, and b are the corresponding normalized results. The calculation formula is as follows:

[0077] (5)

[0078] The combined vegetation index mainly uses color reflectance as a classification standard and is particularly sensitive to green. However, in rural and urban building scenes, there are often a small number of green roofs. If only the combined vegetation index is used as a classification standard for vegetation, such roofs will be treated as vegetation canopies and filtered out. To solve this problem, an RGB color difference constraint is added for green roofs. Although vegetation points and green roofs are both green, the green roof surface is mainly sprayed with artificial synthetic pigments, which is different from the green of naturally grown vegetation. This difference can be described by color difference (ColorDistance), and the formula is as follows:

[0079] (6)

[0080] Where R1, G1, B1 and R2, G2, B2 are the RGB values ​​of point P1 and point P2 respectively. When the green roof seed point is growing, if the color difference between the seed point P1 and the neighboring point P2 is Less than threshold , then the two points are considered to belong to the same roof surface, and region growing continues; if The value exceeds the set threshold , the neighborhood points are identified as vegetation points and regional growth stops.

[0081] Eliminate non-roof point clouds: After the region growing segmentation is completed, the height and area of ​​the roof surface are used to filter and optimize the segmentation results:

[0082] 1) Roof area: Calculate the minimum number of points on the roof based on the point density and the minimum roof area. Remove all small planes with fewer points than the minimum value.

[0083] 2) Height of building roof: defines the minimum roof surface height. When the height of the extracted roof surface point cloud is less than the set height, it is considered that the plane is not a building roof point cloud. The value cannot be used directly as the height. It needs to be combined with the ground point cloud for elevation normalization. The value represents the height from the point to the ground.

[0084] Example 2

[0085] System experiment example:

[0086] A data preprocessing module 10 is used to distinguish ground points from non-ground points using a cloth simulation filtering algorithm;

[0087] The roof surface point cloud extraction module 20 uses a seed point extraction algorithm based on neighborhood information to extract a set of seed points with semantic information about the roof surface. It then performs classification and growth segmentation on vegetation points and non-vegetation points under different growth constraint conditions to initially obtain a roof surface point cloud.

[0088] The data post-processing module 30 performs classification, growth and segmentation to preliminarily obtain the roof surface point cloud; finally, the roof surface point cloud extraction results are filtered and optimized using the height and area values ​​of the roof surface, and ultimately accurate building roof surface point cloud data is obtained.

[0089] In some implementations, this example experiment uses part of the point cloud data in the Estonian terrain dataset. In order to verify the adaptability of the algorithm of this example to different building scenes, three test areas in different scenes are selected, which are respectively recorded as sample1, sample2, and sample3. Figure 7 As shown. Among them, the sample1 measurement area is a rural residential area, which is mainly composed of single small buildings, with dense buildings and diversified structures, and the average building height is about 6m. The vegetation coverage rate is high, and there are many adhesions and occlusions with the buildings. The sample2 measurement area is an urban residential area, which is mainly composed of single medium-sized buildings, with sparse buildings and single structures, and the average building height is about 16m. The vegetation coverage rate is high, but because the buildings are high, there are fewer adhesions and occlusions between vegetation and buildings. The sample3 measurement area is an industrial plant area, which is mainly composed of connected large buildings, and the average building height is about 8m. Because there are many attached roof surfaces and this embodiment experiment selected some point cloud data in the Estonian terrain dataset. In order to verify the adaptability of the algorithm of this embodiment to different building scenes, a total of three test areas in different scenes were selected, which are recorded as sample1, sample2, and sample3, as shown below. Figure 7 As shown. Among them, the sample1 measurement area is a rural residential area, which is mainly composed of small-scale individual buildings. The buildings are dense and the structures are diverse. The average building height is about 6m. The vegetation coverage rate is high, and there are many adhesions and obstructions with the buildings. The sample2 measurement area is an urban residential area, which is mainly composed of medium-sized individual buildings. The buildings are sparse and the structure is simple. The average building height is about 16m. The vegetation coverage rate is high, but because the buildings are high, there are fewer adhesions and obstructions between vegetation and buildings. The sample3 measurement area is an industrial plant area, which is mainly composed of large-scale connected buildings. The average building height is about 8m. Due to the large number of attached roof surfaces and the accuracy of the roof point cloud extraction algorithm, this embodiment uses the roof surface point cloud extracted by manual visual interpretation as the standard reference data, as shown Figure 8shown.

[0090] Experimental parameter setting: As a comprehensive algorithm for roof point cloud extraction, the algorithm in this embodiment involves many algorithm parameters, but most of the parameters are general parameters and will not have a significant impact on the algorithm performance. The parameter settings are as follows: Classification threshold of ground filtering , the number of neighborhood search points for feature calculation , the number of neighborhood search points extracted from the seed point , plane coefficient , the neighborhood search radius of region growing , RGB color difference Other parameters are key parameters and need to be adjusted according to different data. See Table 3 for specific settings.

[0091]

[0092] Experimental results and analysis: The roof surface point cloud extraction was performed on the above three groups of test data. The extraction process and results are shown in the following table. Figure 9-11 shown. Figure 9 It is the non-ground point cloud obtained by cloth simulation filtering. Figure 10 This is the extracted roof surface seed point set. It can be seen that the seed points (red) can accurately fall on the flat area of ​​the building roof, with few false positives or missed positives. This shows that the seed point extraction algorithm of this embodiment has strong robustness. Figure 11 This is the final roof surface point cloud obtained after region growing segmentation and data optimization. It can be seen that the extracted roof surface point cloud is complete and has clear boundaries, indicating that the proposed method can also achieve accurate roof surface segmentation even in the case of lush vegetation interference.

[0093] In order to intuitively reflect the effectiveness of the algorithm in this embodiment, the algorithm extraction results are compared with the reference data. The data comparison is shown in Table 4. Figure 12 As shown. In the figure, the green points (TP) represent correctly extracted roof surface points, the red points (FP) represent false points, and the blue points (FN) represent missed points. The Sample2 measurement area has a good extraction effect due to its simple building structure, without serious missed and false points. In the sample1 and sample3 measurement areas, there are 4 false points (marked with red boxes) and 8 missed points (marked with blue boxes). The main reason for the false points is that the point clouds in these complex areas are very similar to the features of the roof surface and cannot be effectively distinguished. The missed points mainly appear in the point clouds of attached roofs. Such roofs are often filtered out during result optimization due to their small area, which leads to the missed points of such roof surfaces.

[0094] The algorithm of this embodiment faced test data with different building scenes, different vegetation coverage rates, and different building types. Although some misclassification occurred, it still achieved good extraction results overall.

[0095]

[0096] Accuracy evaluation: In order to further verify the extraction performance of the algorithm in this embodiment, the error evaluation criteria mentioned in the literature

[16] were adopted to evaluate the accuracy of the algorithm using four indicators: type I error, type II error, total error and Kappa coefficient.

[0097] (7)

[0098] Where: error1 is the type I error, that is, the error when a roof point is mistakenly classified as a non-roof point; error2 is the type II error, that is, the error when a non-roof point is mistakenly classified as a roof point; error3 is the total error; K is the Kappa coefficient; e, f are the number of reference roof points and non-roof points; g, h are the number of algorithm roof points and non-roof points; a is the intersection of the reference roof point and the algorithm roof point; b is the intersection of the reference roof point and the algorithm non-roof point; c is the intersection of the reference non-roof point and the algorithm roof point; d is the intersection of the reference non-roof point and the algorithm non-roof point; n is the total number of points in the test data;

[0099] The accuracy evaluation of the algorithm in this embodiment is shown in Table 5. The total error is less than 2%, indicating that the algorithm proposed in this embodiment has good extraction accuracy. The overall type I error is larger than the type II error, indicating that the number of roof surface point clouds judged as non-roof surface point clouds is relatively large. On the one hand, because the roof of the building contains many attachments (such as chimneys, etc.), the point clouds of such attachments are marked as roofs during manual visual interpretation and extraction, and such point clouds are curved.

[0100] The rate changes greatly, and it is impossible to use the normal vector angle constraint for effective growth. Therefore, there are many omissions in the point cloud of roof attachments. On the other hand, because some buildings have small-area attached roofs above the exit position, and the topological relationship between such attached roofs and the main building roof is unknown during the point cloud extraction process, such roofs are often regarded as single non-building roofs with too small an area and are eliminated. Therefore, there are many omissions in roofs with too small an area. The Class II error is relatively small, indicating that this embodiment has a better filtering effect on adhered or obstructed vegetation, and solves the problem of misclassification of vegetation points. The Kappa coefficient reached about 97%, indicating that the roof surface point cloud extracted by the algorithm of this embodiment is highly consistent with the reference data.

[0101]

[0102] like Figure 13As shown, an embodiment of the present application provides an electronic device, which includes a memory 101 for storing one or more programs and a processor 102. When the one or more programs are executed by the processor 102, any method of the first aspect described above is implemented.

[0103] The system also includes a communication interface 103. The memory 101, processor 102, and communication interface 103 are electrically connected to each other directly or indirectly to enable data transmission or interaction. For example, these components can be electrically connected to each other via one or more communication buses or signal lines. The memory 101 can be used to store software programs and modules, and the processor 102 executes the software programs and modules stored in the memory 101 to perform various functional applications and data processing. The communication interface 103 can be used to communicate signaling or data with other node devices.

[0104] The memory 101 may be, but is not limited to, a random access memory 101 (RAM), a read-only memory 101 (ROM), a programmable read-only memory 101 (PROM), an erasable programmable read-only memory 101 (EPROM), an electrically erasable programmable read-only memory 101 (EEPROM), etc.

[0105] The processor 102 may be an integrated circuit chip with signal processing capabilities. The processor 102 may be a general-purpose processor 102, including a central processing unit (CPU) or a network processor (NP). It may also be a digital signal processing (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0106] In the embodiments provided in this application, it should be understood that the disclosed methods can also be implemented in other ways. The method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of the methods, methods, and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment, or a portion of code, and the module, program segment, or a portion of code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions.

[0107] In addition, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0108] On the other hand, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When executed by the processor 102, the computer program implements the method described in any of the first aspects above. If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory 101 (ROM), a random access memory 101 (RAM), a magnetic disk, or an optical disk.

[0109] In summary, the method for extracting building roof surface point clouds from airborne lidar data provided in the embodiments of the present application can effectively extract robust roof surface seed points, avoiding serious false extraction and omission phenomena. The segmented roof surface is not only complete but also has clear boundaries, avoiding serious under-segmentation and over-segmentation phenomena. The proposed algorithm has a simple principle and high computational efficiency. The proposed algorithm effectively solves the problem of low extraction accuracy caused by the misclassification of vegetation points. The algorithm can achieve good roof surface point cloud extraction results and has good adaptability to different building scenarios.

[0110] The above are merely preferred embodiments of the present application and are not intended to limit the present application. Persons skilled in the art will readily appreciate that various modifications and variations are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

[0111] It will be apparent to those skilled in the art that the present application is not limited to the details of the exemplary embodiments described above and that the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the present application is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A method for extracting building roof point clouds from airborne laser radar data, characterized in that: include: Use cloth simulation filtering algorithm to distinguish ground points from non-ground points; A seed point extraction algorithm based on neighborhood information is used to extract a set of seed points with semantic information of roof surfaces. Specifically, it includes: extracting constraints based on the neighborhood characteristics of the roof surface point cloud through neighborhood curvature statistics, neighborhood maximum distance and neighborhood gradient analysis; the neighborhood curvature statistics include: obtaining the normal vector and curvature by calculating the eigenvalues ​​and eigenvectors of the neighborhood covariance matrix, and calculating the curvature value of the point cloud; traversing all curvature values, retaining the curvature threshold less than the preset point, obtaining a set of candidate seed points, and then performing a nearest neighbor search on the candidate seed points to achieve plane point extraction; the neighborhood maximum distance includes: eliminating the interference planes with narrow and long characteristics in the plane point cloud extracted by using the neighborhood curvature statistics constraint through the neighborhood maximum distance constraint; the neighborhood gradient analysis includes: distinguishing the roof surface and the building facade based on their different characteristics by using neighborhood gradient information; Classify and grow the vegetation points and non-vegetation points under different growth constraints to obtain the roof surface point cloud. Specifically, the method includes: setting a loose normal vector angle threshold Ensure the integrity of building roof extraction, and then use the combined vegetation index COM and RGB color difference As an additional constraint, vegetation points and non-vegetation points are classified to solve the under-segmentation problem caused by point cloud segmentation; The roof point cloud extraction results are filtered and optimized using the height and area values ​​of the roof surface, and accurate building roof point cloud data is finally obtained.

2. The method for extracting building roof surface point clouds from airborne laser radar data according to claim 1, characterized in that: Also includes: After the region growing segmentation is completed, the height and area values ​​of the roof surface are used to filter and optimize the segmentation results.

3. A building roof point cloud extraction system based on airborne laser radar data, characterized in that: include: Data preprocessing module, used to distinguish ground points from non-ground points using cloth simulation filtering algorithm; The roof surface point cloud extraction module uses a seed point extraction algorithm based on neighborhood information to extract a seed point set with roof surface semantic information; specifically, it includes: according to the neighborhood characteristics of the roof surface point cloud, extraction constraints are performed through neighborhood curvature statistics, neighborhood maximum distance and neighborhood gradient analysis; the neighborhood curvature statistics include: obtaining the normal vector and curvature by calculating the eigenvalues ​​and eigenvectors of the neighborhood covariance matrix, and calculating the point cloud curvature value; traversing all curvature values, retaining the curvature threshold less than the preset point, obtaining a candidate seed point set, and then performing a nearest neighbor search on the candidate seed points to achieve plane point extraction; the neighborhood maximum distance includes: eliminating the interference planes with narrow and long characteristics in the plane point cloud extracted by using the neighborhood curvature statistics constraint through the neighborhood maximum distance constraint; the neighborhood gradient analysis includes: distinguishing the roof surface from the building facade based on the different characteristics of the roof surface; then, classifying and growing the vegetation points and non-vegetation points under different growth constraints to preliminarily obtain the roof surface point cloud; specifically, it includes: setting a loose normal vector angle threshold Ensure the integrity of building roof extraction, and then use the combined vegetation index COM and RGB color difference As an additional constraint, vegetation points and non-vegetation points are classified to solve the under-segmentation problem caused by point cloud segmentation; The data post-processing module performs classification, growth and segmentation to initially obtain the roof surface point cloud; finally, the roof surface height and area values ​​are used to filter and optimize the roof surface point cloud extraction results, and ultimately obtain accurate building roof surface point cloud data.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 2 is implemented.

Images