A method for extracting building contours from ground laser point clouds based on an improved Alpha Shapes algorithm

Abstract

The present invention discloses a method for extracting building contours from ground laser point clouds based on an improved Alpha Shapes algorithm, which is characterized by comprising the following steps: (1) firstly performing building facade segmentation based on a RANSAC algorithm to obtain a plurality of relatively independent building facade point clouds; (2) establishing a Delaunay triangulation network according to the point set of each plane; (3) setting the radius R of the detection circle in the Alpha Shapes algorithm to 1 to 2 times the average point spacing; (4) if the length of a certain side in a triangle is greater than 2R, the triangle is deleted; (5) judging each side of the triangle: if a circle with a radius of R passing through two points of a certain side contains other points, the triangle is deleted; (6) finding the edge of the triangulation network on the obtained triangulation network; (7) merging point cloud patches to obtain a complete building contour line. The present invention has the advantages of high efficiency and good stability.

Description

Technical Field

[0001] The invention relates to the field of geographic information technology, and in particular to a method for extracting building contours from laser point clouds based on an improved Alpha Shapes algorithm. Background Art

[0002] As a new spatial data acquisition method, the ground-based three-dimensional laser scanning system takes into account the high-precision requirements of distance and spatial resolution. It can quickly, accurately, continuously and automatically obtain high-precision three-dimensional point clouds about the facades of buildings. The contour information of buildings is an important basis for building extraction and three-dimensional model reconstruction. It has been widely used in urban basic information database update, target recognition, disaster prediction, change detection, real estate and other fields. At present, the general method of tracking the plane contour of buildings based on laser point cloud data is to use an interpolation algorithm to assign the elevation value of the three-dimensional point cloud to the image grayscale value, further process it to obtain a regularized nDSM, and finally extract the building contour line by image segmentation, edge detection or combining high-resolution images. The problem with these methods is that the tracked edges are rough boundaries of discrete point sets and have low accuracy.

[0003] Some scholars have also studied methods of directly extracting contours from discrete point sets. For example, Huang Xianfeng et al. proposed an edge tracking algorithm based on planar discrete points. This algorithm uses the side length ratio as a constraint condition, which reduces the dependence of the algorithm parameters on point density, thereby improving the algorithm's adaptability to edge extraction of slender features or unevenly distributed point sets. However, improper setting of the constraint threshold can easily cause edge transition shrinkage.

[0004] The Alpha Shapes algorithm was first proposed by Edelsbrunner et al. Later, many scholars improved it and applied it to the field of laser point cloud data processing. The algorithm is theoretically sound and efficient, and can also handle more complex building contour extraction problems. The main disadvantage is that it is not suitable for unevenly distributed data, and the selection of algorithm parameters is also difficult. When the radius α of the detection circle is large, the detection circle rolls outside the point set S, and the trace of the external rolling is the boundary contour of the point set. When the radius α value is small, the detection circle rolls inside the point set S. When the α value is small enough, every point in the point set is a boundary point. The value of the circle radius α is closely related to the fineness of the detected building contour. When the radius is relatively small, the detected contour is finer, and when the radius is relatively large, the detected contour is slightly rough. Therefore, the existing method needs to be further improved. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a laser point cloud building contour extraction method based on an improved Alpha Shapes algorithm, which is used to solve the problem of low efficiency in contour extraction of existing buildings.

[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0007] The method for extracting building contours from ground laser point clouds based on the improved Alpha Shapes algorithm includes the following steps:

[0008] Step 1: First, segment the building facade based on the RANSAC algorithm to obtain multiple relatively independent building facade point clouds;

[0009] Step 2: Establish a Delaunay triangulation based on the point set of each plane;

[0010] Step 3: Set the radius R of the detection circle in the Alpha shapes algorithm, which is generally set to 1 to 2 times the average point spacing;

[0011] Step 4: If the length of a side of the triangle is greater than 2R, delete the triangle;

[0012] Step 5: Judge each side of the triangle: if a circle with a radius of R passing through two points of a certain side contains other points, delete the triangle;

[0013] Step 6: Find the edge of the triangulated network on the obtained triangulated network;

[0014] Step 7: Merge the point cloud patches to obtain the complete building outline.

[0015] Preferably, the specific method for performing building facade segmentation based on the RANSAC algorithm adopted in step 1 is:

[0016] Step 1.1: Set the point set Q to be empty, the optimal point set Q_best to be empty, and the number of iterations to 1;

[0017] Step 1.2: Randomly select three points and calculate the plane parameters, which are recorded as model M;

[0018] Step 1.3: Calculate the deviation between all points in the point set and the model, and add the points with a deviation less than 0.5 to the point set Q;

[0019] Step 1.4: If the number of elements in the current point set Q is greater than the optimal point set Q_best, update Q_best = Q and update the number of iterations at the same time; if the number of iterations is greater than 10000, exit, and the optimal point set Q_best is a point cloud plane segmented from the building; otherwise, add 1 to the number of iterations, set Q to an empty set, and repeat the above steps 1.2 to 1.4.

[0020] Step 1.5: Repeat steps 1.2 to 1.4 above.

[0021] Preferably, the specific method of establishing the Delaunay triangulation in step 2 is:

[0022] Step 2.1: Construct a hypertriangle or hyperpolygon of the plane point set p and use it as the convex closure of the point set;

[0023] Step 2.2: Insert a point from the point set p into the triangulated network. During the insertion process, it is necessary to satisfy the empty circle rule of the Delaunary triangulated network. If there is a circumscribed circle of triangle Ti in the triangulated network that contains the insertion point, it will be deleted;

[0024] Step 2.3: Construct a new triangulated network based on the insertion point and the non-common edges of triangle Ti, and iterate in this way until the point set p is empty, then the algorithm stops.

[0025] The technical concept of the present invention is to propose a method for extracting building contours from laser point clouds based on an improved Alpha Shapes algorithm. First, the facades of the building are segmented by adjusting parameters; then the contours of each facet are extracted; and finally, the facade facets are merged to obtain the contour of the entire building. The present invention can extract the building contour without relying on prior information.

[0026] The invention has the advantages of being able to extract the building contour without any other prior information, being efficient, being stable, etc., and providing good technical support for the three-dimensional modeling of buildings. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is the overall flow chart of the present invention.

[0028] Figure 2 This is the overall effect diagram of the building facade segmentation.

[0029] Figure 3a to 3c These are the three different facade patches of the building.

[0030] Figures 4a to 4c for Figure 3a to 3c Contour extraction diagram corresponding to each facade patch of the building.

[0031] Figure 5 Extract the overall outline of the building. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The specific embodiments described here are only used to explain the present invention and are not used to limit the present invention.

[0033] The method for extracting building contours from ground laser point clouds based on the improved Alpha Shapes algorithm includes the following steps:

[0034] Step 1: First, segment the building facade based on the RANSAC algorithm to obtain multiple relatively independent building facade point clouds;

[0035] Step 1.1: Set the point set Q to be empty, the optimal point set Q_best to be empty, and the number of iterations to 1;

[0036] Step 1.2: Randomly select three points and calculate the plane parameters, which are recorded as model M;

[0037] Step 1.3: Calculate the deviation between all points in the point set and the model, and add the points with a deviation less than 0.5 to the point set Q;

[0038] Step 1.4: If the number of elements in the current point set Q is greater than the optimal point set Q_best, update Q_best = Q and update the number of iterations at the same time; if the number of iterations is greater than 10000, exit, and the optimal point set Q_best is a point cloud plane segmented from the building; otherwise, add 1 to the number of iterations, set Q to an empty set, and repeat the above steps 1.2 to 1.4.

[0039] This example is a single building scanned by the LiBackpack D50 backpack laser radar scanning system. When using the RANSAC algorithm to extract the facade, the plane deviation threshold is set in advance to complete the facade segmentation of this example. The overall effect of the building facade segmentation is shown in the figure. Figure 2 As shown, a single facade patch is Figure 3a to 3c shown.

[0040] Step 2: Establish a Delaunay triangulation based on the point set of each plane;

[0041] Step 2.1: Construct a hypertriangle or other hyperpolygon of the plane point set p and use it as the convex closure of the point set;

[0042] Step 2.2: Insert a point from the point set p into the triangulated network. During the insertion process, it is necessary to satisfy the empty circle rule of the Delaunary triangulated network. If there is a circumscribed circle of triangle Ti in the triangulated network that contains the insertion point, it will be deleted;

[0043] Step 2.3: Construct a new triangulated network based on the insertion point and the non-common edges of triangle Ti, and iterate in this way until the point set p is empty, then the algorithm stops.

[0044] Step 3: Set the radius R of the detection circle in the Alpha shapes algorithm, which is generally set to 1 to 2 times the average point spacing;

[0045] Step 4: If the length of a side of the triangle is greater than 2R, delete the triangle;

[0046] Step 5: Judge each side of the triangle: if a circle with a radius of R passing through two points of a certain side contains other points, delete the triangle;

[0047] Step 6: Connect all points on the obtained triangulated network point by point to obtain the contour line of the building facade. The contour line extraction diagram corresponding to each facade face of the building in this embodiment is as follows: Figures 4a to 4c shown.

[0048] Step 7: Merge the point cloud patches to obtain the complete building outline. Point cloud merging is the visualization loading of multiple point clouds in the same window. In this embodiment, the overall building outline extraction diagram is as follows: Figure 5 shown.

[0049] The present invention makes full use of the characteristics of the building itself and can realize the contour extraction of complex, noisy three-dimensional point cloud buildings. The present invention does not require prior information and has obvious advantages in terms of accuracy and robustness.

[0050] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Ordinary technicians in the field can modify or replace the technical solution of the present invention with equivalents without departing from the principle and scope of the present invention.

Claims

1. A method for extracting building contours from ground laser point clouds based on an improved Alpha Shapes algorithm, characterized in that: The steps include: Step 1: First, segment the building facade based on the RANSAC algorithm to obtain multiple relatively independent building facade point clouds; Step 2: Establish a Delaunay triangulation based on the point set of each plane; Step 3: Set the radius R of the detection circle in the Alpha shapes algorithm to 1 to 2 times the average point spacing; Step 4: If the length of a side of the triangle is greater than 2R, delete the triangle; Step 5: Check each side of the triangle: if a circle with a radius of R passing through two points of a side contains other points except these two points, delete the triangle; Step 6: Find the edge of the triangulated network on the obtained triangulated network; Step 7: Merge the point cloud patches to obtain the complete building outline; The building facade segmentation based on the RANSAC algorithm described in step 1 specifically includes: Step 1.1: Set the point set Q to be empty, the optimal point set Q_best to be empty, and the number of iterations to 1; Step 1.2: Randomly select three points and calculate the plane parameters, which are recorded as model M; Step 1.3: Calculate the deviation between all points in the point set and the model, and add the points with a deviation less than 0.5 to the point set Q; Step 1.4: If the number of elements in the current point set Q is greater than the optimal point set Q_best, then update Q_best = Q and update the number of iterations at the same time; if the number of iterations is greater than 10000, then exit, and the optimal point set Q_best is a point cloud plane segmented from the building; otherwise, the number of iterations is increased by 1, and Q is set to an empty set; Step 1.5: Repeat steps 1.2 to 1.4 above.

2. The method for extracting building contours from ground laser point clouds based on an improved Alpha Shapes algorithm according to claim 1, characterized in that: The establishment of the Delaunay triangulation described in step 2 specifically includes: Step 2.1: Construct a hypertriangle or hyperpolygon of the plane point set p and use it as the convex closure of the point set; Step 2.2: Insert a point from the point set p into the triangulated network. During the insertion process, it is necessary to satisfy the empty circle rule of the Delaunary triangulated network. If there is a circumscribed circle of triangle Ti in the triangulated network that contains the insertion point, it will be deleted; Step 2.3: Construct a new triangulated network based on the insertion point and the non-common edges of triangle Ti, and iterate in this way until the point set p is empty, then the algorithm stops.

Images