A method, system and medium for automatically and quickly filtering vegetation from point cloud in complex mountainous areas

Abstract

The application discloses a kind of complex mountainous area's point cloud vegetation automatic fast filtering method, system and medium;It relates to data processing technical field;The point cloud data of target area is acquired;The point cloud data is preliminarily filtered;The point cloud data after preliminary filtering is clustered based on DBSCAN algorithm, and finally the point cloud data after filtering of target area is obtained;The present application is improved based on the existing data processing technology, and block area restriction is introduced, combined with height restriction, slope restriction and neighborhood restriction condition, the point cloud data is preliminarily filtered first, then the point cloud data after preliminary filtering is clustered based on DBSCAN algorithm, to automatically, reliable way filters and separates vegetation in point cloud, effectively overcome the problems such as noise, shielding and complex geometry, realize more accurate, complete point cloud data reconstruction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a point cloud vegetation automatic rapid filtering method and system for complex mountainous areas and a medium. BACKGROUND

[0002] Currently, there are some methods for filtering vegetation point clouds. One common method is Cloth Simulation Filtering (CSF), which is an algorithm based on a simple physical process. The basic idea is to place a virtual piece of cloth above the terrain. Under the action of gravity, the cloth will descend. Assuming that the cloth is very soft, the cloth will adhere to the surface of the ground object, and the final shape of the cloth constitutes a digital surface model (DSM). If the terrain is inverted in advance and the cloth has a certain hardness, the final shape of the cloth constitutes a high-precision digital elevation model. Another method is a filtering method based on morphological operations. Progressive mathematical morphological filtering mainly uses the opening operation in mathematical morphology. By performing erosion followed by dilation, small targets can be removed. The algorithm first performs a regular gridding operation on the original LiDAR point cloud. By comparing the height difference between the corresponding grid points before and after the opening operation with the set threshold, points greater than the threshold are removed. The filter window is expanded in a linear or exponential manner for iterative calculation until the filter window reaches the set maximum value.

[0003] These existing methods have some shortcomings in solving the problem of vegetation point cloud filtering. Specifically:

[0004] 1. The Cloth Simulation Filtering (CSF) method can obtain a digital surface model by simulating a physical process, but it relies on the assumed cloth properties and hardness parameters, which are sensitive to parameter selection and prone to errors.

[0005] 2. The filtering method based on morphological operations can better remove small targets through the opening operation, but when there are large noise or occlusions in the vegetation point cloud, it may affect the accuracy of the filtering result.

[0006] 3. The progressive mathematical morphological filtering method can partially solve the information loss problem caused by a small filter window by iteratively expanding the filter window for calculation. However, when dealing with complex vegetation scenes, a larger filter window may need to be set, increasing the computational complexity and time consumption.

[0007] These methods may have certain limitations when dealing with different vegetation types, noise interference, and occlusions. Therefore, it is necessary to further improve the accuracy, robustness, and universality of existing methods to meet more extensive application requirements. SUMMARY

[0008] The technical problem to be solved by the present application is that current point cloud processing technology faces a series of difficulties and limitations when processing vegetation point cloud data, such as noise, occlusion and complex geometric structure; these problems lead to the inaccuracy and incompleteness of vegetation point cloud data, which in turn affects the accuracy and effect of related application fields. The present application aims to provide a point cloud vegetation automatic fast filtering method, system and medium in complex mountainous areas, which is based on the improvement of the existing data processing technology, and the point cloud data is first filtered, and then the DBSCAN algorithm is used to cluster the preliminary filtered point cloud data, so as to automatically and reliably filter and segment the vegetation in the point cloud, effectively overcoming the problems of noise, occlusion and complex geometric structure, and realizing more accurate and complete point cloud data reconstruction.

[0009] The present application is realized by the following technical scheme:

[0010] The present application provides a point cloud vegetation automatic fast filtering method in complex mountainous areas, comprising:

[0011] Obtaining point cloud data of a target area;

[0012] Preliminary filtering of point cloud data: dividing the point cloud data into multiple block areas, and limiting the height, slope and neighborhood of each block area;

[0013] Clustering the preliminary filtered point cloud data based on the DBSCAN algorithm to obtain the filtered point cloud data of the target area.

[0014] The working principle of the present application is that current point cloud processing technology faces a series of difficulties and limitations when processing vegetation point cloud data, such as noise, occlusion and complex geometric structure; these problems lead to the inaccuracy and incompleteness of vegetation point cloud data, which in turn affects the accuracy and effect of related application fields. The present application aims to provide a point cloud vegetation automatic fast filtering method, system and medium in complex mountainous areas, which is based on the improvement of the existing data processing technology, and the point cloud data is first filtered, and then the DBSCAN algorithm is used to cluster the preliminary filtered point cloud data, so as to automatically and reliably filter and segment the vegetation in the point cloud, effectively overcoming the problems of noise, occlusion and complex geometric structure, and realizing more accurate and complete point cloud data reconstruction.

[0015] Further optimization scheme is that the point cloud data is divided into multiple block areas, comprising the method:

[0016] Projecting the point cloud data of the target area into two-dimensional plane data;

[0017] The two-dimensional plane data is divided into mxn block regions based on the block coefficient bc:

[0018]

[0019]

[0020] Wherein, X max , X min respectively represent the maximum value of X coordinate and the minimum value of X coordinate in the two-dimensional plane data, Y max , Y min respectively represent the maximum value of Y coordinate and the minimum value of Y coordinate in the two-dimensional plane data, and the block coefficient bc represents the area of the block.

[0021] The further optimization scheme is that the height limit, slope limit and neighborhood limit are respectively performed on each block region, including the method:

[0022] S1, the minimum elevation difference Δz between the current block region and the neighborhood block is calculated:

[0023] Δz = P ij z-min(P i-1j-1 z, P i-1j z, P i-1j+1 z, P ij-1 z, P ij+1 z, P i+1j-1 z, P ij+1 z, P i+1j+1 z)

[0024] Wherein, P ij is the point with the minimum elevation in all point clouds of the current block R ij , P ij x, P ij y, P ij z are the coordinates of P ij corresponding to X axis, Y axis and Z axis; 1≤i≤m,1≤j≤n;

[0025] If Δz≤NT, then go to step S2, otherwise, all point clouds in the current block R ij are regarded as vegetation points;

[0026] S2, the height A ij and the slope slope of the point cloud A in the current block R z are obtained; the height A z and the slope slope of the point cloud A are respectively limited in height and slope, and the point cloud in the current block R ij which satisfies the height limit and the slope limit is added to the non-vegetation point set;

[0027] S3, repeating steps S1-S2 to obtain the preliminary filtered point cloud data of each block.

[0028] Further, the height restriction comprises the following method:

[0029] determining whether the height A z satisfies A z -P ij z≤HT, where HT represents a height threshold; if yes, the point cloud A satisfies the height restriction.

[0030] Further, the slope calculation method of the point cloud A comprises the following method:

[0031] The slope of the point cloud A is calculated according to the following formula:

[0032]

[0033] dz=|A z -P ij z|

[0034]

[0035] Further, the slope restriction comprises the following method:

[0036] determining whether the slope of the point cloud A satisfies slope≤ST, where ST represents a slope threshold; if yes, the point cloud A satisfies the slope restriction.

[0037] Further, the clustering of the preliminary filtered point cloud data based on the DBSCAN algorithm comprises the following method:

[0038] The preliminary filtered point cloud data is clustered according to density based on the DBSCAN algorithm.

[0039] Further, the clustering of the preliminary filtered point cloud data based on the DBSCAN algorithm further comprises the following method: when the preliminary filtered point cloud data is clustered based on the DBSCAN algorithm, the number of point clouds in each cluster is obtained; if the number of point clouds in the current cluster is less than a number threshold, all point clouds in the cluster are discarded.

[0040] The present scheme also provides a point cloud vegetation automatic rapid filtering system for complex mountainous areas, which is used to implement the point cloud vegetation automatic rapid filtering method for complex mountainous areas, and comprises the following:

[0041] A collection module is configured to obtain point cloud data of a target region.

[0042] A preliminary filtering module is configured to preliminarily filter the point cloud data, that is, to divide the point cloud data into a plurality of block regions, and to perform height limitation, slope limitation and neighborhood limitation on each block region respectively;

[0043] A clustering filtering module is configured to cluster the preliminarily filtered point cloud data based on a DBSCAN algorithm, so as to finally obtain the filtered point cloud data of the target region.

[0044] The present application also provides a computer readable medium having a computer program stored thereon, wherein the computer program is executable by a processor to implement the point cloud vegetation automatic fast filtering method for complex mountainous areas.

[0045] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0046] The present application provides a point cloud vegetation automatic fast filtering method for complex mountainous areas, a system and a medium. Based on the existing data processing technology, the method is improved by introducing block region limitation, combining height limitation, slope limitation and neighborhood limitation conditions, preliminarily filtering the point cloud data, and clustering the preliminarily filtered point cloud data based on a DBSCAN algorithm, so as to automatically and reliably filter and segment the vegetation in the point cloud, effectively overcome the problems of noise, occlusion and complex geometric structure, and realize more accurate and complete point cloud data reconstruction. BRIEF DESCRIPTION OF DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the example embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope, and for those skilled in the art, other related drawings can also be obtained without creative labor. In the drawings:

[0048] Figure 1 It is a flowchart of the point cloud vegetation automatic fast filtering method for complex mountainous areas;

[0049] Figure 2 It is a schematic diagram of the point cloud data filtering process;

[0050] Figure 3 It is a schematic diagram of the original point cloud;

[0051] Figure 4 It is a schematic diagram of the point cloud block division;

[0052] Figure 5 It is a schematic diagram of the block filtering process;

[0053] Figure 6 It is a schematic diagram of the ground point set after preliminary filtering;

[0054] Figure 7 A schematic diagram of a ground point set filtered by DBSCAN. DETAILED DESCRIPTION

[0055] In order to make the objects, technical solutions and advantages of the present application clearer, further detailed description will be given to the present application in combination with embodiments and drawings, the illustrative embodiments of the present application and the description thereof are only used to explain the present application, and do not limit the present application.

[0056] Current point cloud processing technology faces a series of difficulties and limitations when processing vegetation point cloud data, such as noise, occlusion and complex geometry; these problems lead to the inaccuracy and incompleteness of vegetation point cloud data, which further affects the accuracy and effect of related application fields. In view of this, the present application provides the following embodiments to solve the above technical problems:

[0057] Embodiment 1

[0058] The embodiment provides a method for automatically and quickly filtering point cloud vegetation in complex mountainous areas, as shown in Figure 1 and Figure 2 , comprising:

[0059] Step 1: Obtain point cloud data of the target area;

[0060] Step 2: Preliminary filtering of point cloud data: divide the point cloud data into multiple block areas, and limit the height, slope and neighborhood of each block area respectively;

[0061] Step 3: Clustering the preliminary filtered point cloud data based on DBSCAN algorithm, and finally obtaining the filtered point cloud data of the target area.

[0062] The point cloud data is divided into multiple block areas, including the method:

[0063] Projecting the point cloud data of the target area into two-dimensional plane data;

[0064] For mountain point cloud data, direct vegetation filtering, algorithm has difficulty. Choose to use the block idea, project the three-dimensional point cloud data to the X-Y plane, divide it into m×n block areas using the block coefficient bc:

[0065]

[0066]

[0067] Wherein, X max , X min respectively represent the maximum value of X coordinate and the minimum value of X coordinate in two-dimensional plane data, Y max , Ymin respectively represent the maximum value of Y coordinate and the minimum value of Y coordinate in two-dimensional plane data, and the block coefficient bc represents the area of the block; when bc = 1, it represents that the area of the block is 1*1 square meter.

[0068] The height limit, slope limit and neighborhood limit for each block region respectively include the following methods:

[0069] S1, calculating the minimum elevation difference Δz between the current block and the neighborhood block:

[0070] Δz = P ij z-min(P i-1j-1 z, P i-1j z, P i-1j+1 z, P ij-1 z, P ij+1 z, P i+1j-1 z, P ij+1 z, P i+1j+1 z)

[0071] wherein, P ij is the point with the minimum elevation in all point clouds of the current block R ij , P ij x, P ij y, P ij z are the coordinates of P ij corresponding to the X axis, Y axis and Z axis; 1≤i≤m, 1≤j≤n;

[0072] If Δz≤NT, then go to step S2, otherwise, all point clouds in the current block R ij are regarded as vegetation points;

[0073] This step is mainly for high trees or power transmission lines, which have high height and are screened out by height limit.

[0074] S2, obtaining the height A ij and slope of the point cloud A in the current block R z ; performing height limit and slope limit on the height A z and slope of the point cloud A respectively, and adding the point cloud in the current block R ij that meets the height limit and slope limit to the non-vegetation point set;

[0075] The height limit includes the following methods:

[0076] judging whether the height A z meets A z -P ij z≤HT, wherein HT represents the height threshold; if yes, the point cloud A meets the height limit. The height limit is mainly for high trees or power transmission lines.

[0077] The slope slope of the point cloud A is calculated by the following method:

[0078] The slope slope of the point cloud A is calculated by the following formula:

[0079]

[0080] dz = |A z -P ij z|

[0081]

[0082] The slope limit includes the following method:

[0083] It is judged whether the slope slope of the point cloud A satisfies slope ≤ ST, and ST represents a slope threshold; if yes, the point cloud A satisfies the slope limit.

[0084] The neighborhood limit is used to judge whether there is an obvious fault between the current block area point cloud and the surrounding; the vegetation includes tall trees, shrubs and the like, and the power poles are similar to the trees, and their obvious features are high height, and the height limit can well filter such point clouds; the LiDAR elevation minimum point is often a ground point, and all the elevation minimum points in the block area are taken as the reference; meanwhile, the trees and power poles in the vegetation tend to be vertical, and the slope is large, and the slope limit can well filter such point clouds.

[0085] S3, the steps S1-S2 are cycled to obtain the point cloud data after preliminary filtering.

[0086] The DBSCAN algorithm is used to cluster the point cloud data after preliminary filtering, including the following method:

[0087] The DBSCAN algorithm is used to cluster the point cloud data after preliminary filtering according to the density.

[0088] The DBSCAN algorithm is used to cluster the point cloud data after preliminary filtering, and further includes the following method: when the DBSCAN algorithm is used to cluster the point cloud data after preliminary filtering, the number of point clouds in each cluster category is obtained, and if the number of point clouds in the current category is less than a number threshold, all the point clouds in the category are discarded. When clustering, since the discrete points are few and unevenly distributed, the point clouds with a category less than the number threshold can be discarded to obtain the point cloud after final filtering.

[0089] The preliminary filtering often has discrete points, and the classical DBSCAN algorithm is used for clustering to solve the problem of discrete points. DBSCAN can cluster according to density. When clustering, since the discrete points are few and unevenly distributed, the point clouds with a category less than a certain number can be discarded to obtain the point cloud after final filtering.

[0090] Embodiment 2

[0091] The embodiment provides a point cloud vegetation automatic rapid filtering system for complex mountainous areas, which is used for implementing the point cloud vegetation automatic rapid filtering method for complex mountainous areas in the embodiment 1, and comprises the following steps of:

[0092] A collection module is configured to acquire point cloud data of a target region.

[0093] A preliminary filtering module is configured to perform preliminary filtering on the point cloud data, and the point cloud data is divided into a plurality of block regions, and height limitation, slope limitation and neighborhood limitation are performed on each block region.

[0094] A clustering filtering module is configured to perform clustering on the point cloud data after the preliminary filtering based on a DBSCAN algorithm, and finally obtain filtered point cloud data of the target region.

[0095] Embodiment 3

[0096] The embodiment provides a computer readable medium, and a computer program is stored on the computer readable medium, and the computer program is executed by a processor to implement the point cloud vegetation automatic rapid filtering method for complex mountainous areas in the embodiment 1.

[0097] Embodiment 4

[0098] The embodiment takes a region as an example, and a point cloud image of the region is as shown in Figure 3 .

[0099] The first step is point cloud blocking, and the diagram is a perspective view segmentation, which is divided into 6x5 blocks. Figure 4

[0100] The second step is to filter each block region, including neighborhood, height and slope limitation. Taking one block region as an example, the neighborhood threshold is set to 10 meters, the height threshold is set to 3 meters, and the slope threshold is set to 45 degrees, as shown in Figure 5 , the left block region is the original image, the middle block region is the image after filtering the vegetation, and the right block region is the vegetation image.

[0101] The third step is to combine the filtering results of each block region and perform DBSCAN filtering. As shown in Figure 6 , the preliminary filtering result contains some discrete points, which are the antennas in the original image. As shown in Figure 7 , the result after DBSCAN filtering.

[0102] In the block limitation and block combination, multi-process is used for acceleration to achieve rapid filtering of point cloud vegetation. Taking the point cloud of the region as an example, the point cloud file is 179 MB, and the number of point clouds is 9395956. The multi-process calculation takes about 3-4 minutes, and the single-process calculation takes as long as 20 hours.

[0103] ​(1) Using the block idea to project three-dimensional point cloud data to the plane, and using block coefficients for processing.

[0104] (2) Height restriction, slope restriction and neighborhood restriction are performed within the block to filter and identify vegetation point cloud.

[0105] (3) Using DBSCAN algorithm for discrete point filtering and point cloud clustering.

[0106] The block area restriction provided by the embodiment: by defining the block size and calculating the height, slope and other thresholds, the vegetation point cloud in the rugged area is effectively filtered.

[0107] The neighborhood restriction provided by the embodiment: by judging the elevation difference of the block point cloud and the surrounding area, obvious faults are identified, and these point clouds are excluded in subsequent processing.

[0108] The calculation process provided by the embodiment: by blocking, each block is further restricted according to the neighborhood, height and slope, and then the filtered point cloud is obtained by merging, and finally DBSCAN clustering is used for filtering.

[0109] The present application introduces block area restriction, and combines height, slope and neighborhood restrictions to comprehensively process and filter point cloud data in rugged areas. At the same time, DBSCAN algorithm is used for clustering to further optimize the final point cloud filtering effect.

[0110] The above specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for automatic and rapid filtering of point cloud vegetation in complex mountainous areas, characterized in that, include: Acquire point cloud data for the target area; Preliminary filtering of point cloud data: Divide the point cloud data into multiple blocks, and impose height, slope and neighborhood restrictions on each block. The point cloud data after initial filtering is clustered based on the DBSCAN algorithm to finally obtain the point cloud data after filtering the target area. The method for dividing point cloud data into multiple blocks includes: Project the point cloud data of the target area into two-dimensional planar data; Based on the partitioning coefficient bc, the two-dimensional planar data is divided into m x n blocks: ； ； in, These represent the maximum and minimum X-coordinate values ​​in the two-dimensional planar data, respectively. These represent the maximum and minimum Y-coordinate values ​​in the two-dimensional planar data, respectively, and the block coefficient bc represents the area of ​​the block. The method for applying height, slope, and neighborhood restrictions to each block region includes: S1, calculate the minimum elevation difference between the current block and neighboring blocks. : ； in, For the current block The point with the minimum elevation in the entire point cloud. for The coordinates corresponding to the X, Y, and Z axes; ; like If so, proceed to step S2; otherwise, move to the current block. All point clouds within the area are considered vegetation points; S2, Get the current block Height of interior point cloud A and slope; height of point cloud A Apply height and slope limits to the current block. Point clouds that meet the height and slope constraints are added to the set of non-vegetated points; S3, repeat steps S1-S2, traversing each block to obtain the pre-filtered point cloud data.

2. The method for automatic and rapid filtering of point cloud vegetation in complex mountainous areas according to claim 1, characterized in that, The height restriction includes a method: Determine height Does it meet the requirements? HT represents the height threshold; if so, then point cloud A satisfies the height limit.

3. The method for automatic and rapid filtering of point cloud vegetation in complex mountainous areas according to claim 1, characterized in that, The methods for calculating the slope of point cloud A include: The slope of point cloud A is calculated using the following formula: ； ； 。 4. The method for automatic and rapid filtering of point cloud vegetation in complex mountainous areas according to claim 3, characterized in that, The slope limitation includes the following methods: Determine if the slope of point cloud A satisfies ST represents the slope threshold; if so, then point cloud A satisfies the slope constraint.

5. The method for automatic and rapid filtering of point cloud vegetation in complex mountainous areas according to claim 1, characterized in that, The method for clustering the initially filtered point cloud data based on the DBSCAN algorithm includes: The DBSCAN algorithm is used to cluster the initially filtered point cloud data according to density.

6. The method for automatic and rapid filtering of point cloud vegetation in complex mountainous areas according to claim 5, characterized in that, The method for clustering the initially filtered point cloud data based on the DBSCAN algorithm also includes: when clustering the initially filtered point cloud data based on the DBSCAN algorithm, obtaining the number of point clouds in each cluster category; if the number of point clouds in the current category is less than the number threshold, then discarding all point clouds in that category.

7. An automatic and rapid point cloud vegetation filtering system for complex mountainous areas, characterized in that, The method for automatically and rapidly filtering point cloud vegetation in complex mountainous areas as described in any one of claims 1-6 includes: The acquisition module is used to acquire point cloud data of the target area; The preliminary filtering module is used to perform preliminary filtering on point cloud data: it divides the point cloud data into multiple blocks and applies height restrictions, slope restrictions, and neighborhood restrictions to each block. The clustering and filtering module is used to cluster the initially filtered point cloud data based on the DBSCAN algorithm, and finally obtain the filtered point cloud data of the target region.

8. A computer-readable medium having a computer program stored thereon, characterized in that, The computer program, when executed by a processor, can implement a method for automatically and rapidly filtering point cloud vegetation in complex mountainous areas as described in any one of claims 1-6.

Images