CN113627353B - Method for classifying ground points in point cloud data - Google Patents
Method for classifying ground points in point cloud data Download PDFInfo
- Publication number
- CN113627353B CN113627353B CN202110925166.4A CN202110925166A CN113627353B CN 113627353 B CN113627353 B CN 113627353B CN 202110925166 A CN202110925166 A CN 202110925166A CN 113627353 B CN113627353 B CN 113627353B
- Authority
- CN
- China
- Prior art keywords
- grid
- points
- point
- ground
- seed
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/20—Finite element generation, e.g. wire-frame surface description, tesselation
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Length Measuring Devices With Unspecified Measuring Means (AREA)
Abstract
The invention relates to the technical field of data classification, in particular to a method for classifying ground points in point cloud data, which screens grids by meshing the point cloud data, screens rough classification ground points according to the position relationship between the points in the screened grids and the lowest points of the grids, meshes the rough classification ground points, screens fine classification ground points according to the position relationship between the points in the rough classification ground point grids and the average points of the grids, and solves the problems of large calculation amount and low precision of the existing ground point classification.
Description
Technical Field
The invention relates to the technical field of data classification, in particular to a method for classifying ground points in point cloud data.
Background
The point cloud data is a set of points with three-dimensional coordinates, and at present, the basic idea of the method for classifying the ground points by the point cloud data is as follows: a triangle is firstly established on the lowest elevation point of a survey area, then a triangle meeting a certain condition is established upwards on the basis of the lowest triangle according to certain parameters, and the vertex of the triangle is a ground point. Thus, the ground points are classified through the process of repeatedly building triangles, and the classification method has the following problems:
1. due to the huge data volume of the point cloud, the repeated establishment of the triangle has higher requirement on a computer and lower calculation speed;
2. in places with complex terrain, the method still has some errors and needs manual intervention;
3. because of the existence of noise points and the error of the original point cloud, the lowest elevation point is directly used as a seed point to construct a triangular net, and the accuracy of ground points can be reduced.
Disclosure of Invention
The technical problem solved by the invention is as follows: the method for classifying the ground points in the point cloud data solves the problems of large ground point classification calculation amount and low accuracy in the prior art.
The invention adopts the technical scheme for solving the technical problems that: the method for classifying the ground points in the point cloud data comprises the following steps:
s01, meshing the point cloud data;
s02, taking the lowest point in the grid as a first seed point of the grid;
s03, screening the grid according to the position relation between the points in the grid and the first seed points to obtain the screened grid;
s04, obtaining all roughly classified ground points according to the position relation between the points in the screened grid and the corresponding first seed points;
s05, performing grid-netting on all the roughly classified ground points;
s06, calculating the average point of all points in the roughly classified ground point grid, wherein the horizontal coordinate of the average point is the average value of the horizontal coordinates of all points in the grid, the vertical coordinate is the average value of the vertical coordinates of all points in the grid, and the vertical coordinate is the average value of the vertical coordinates of all points in the grid;
s07, taking the average point as a second seed point of the grid;
and S08, obtaining all the finely classified ground points according to the grids and the second seed points.
Further, in step S01, the gridding the point cloud data includes:
taking the size of the maximum building as the size of the grid in the point cloud data containing the building;
and taking the first preset value as the size of the grid in the point cloud data without the building.
Further, in step S03, the method for obtaining the filtered grid includes the following steps:
s301, calculating an average value of the position relationship between each point in each grid and the first seed point of the grid, and obtaining the grid and the corresponding average value;
s302, selecting a grid, if the grid meets the following conditions: if the average value of the grid is smaller than the average value of the adjacent grids and the difference value is smaller than the first threshold value, the grid is the screened grid;
and S303, sequentially selecting each grid to obtain the screened grids.
Further, in step S04, the method for obtaining rough classification ground points includes the following steps:
s401, selecting a screened grid, and if the position relation between a point in the grid and a first seed point corresponding to the point is smaller than a second threshold, determining that the point is a roughly classified ground point, and obtaining all roughly classified ground points in the grid;
s402, sequentially selecting all the screened grids to obtain all the roughly classified ground points.
Further, in step S05, the step of gridding all the roughly classified ground points includes: and taking the second preset value as the size of the grid.
Further, in step S08, the method for obtaining the fine classification ground points includes the following steps:
s801, selecting a grid with the average point as a second seed point, wherein if the position relation between the point in the grid and the second seed point corresponding to the point is smaller than a third threshold value, the point is a fine classification ground point, and all fine classification ground points in the grid are obtained;
s802, sequentially selecting all grids taking the average points as second seed points to obtain all the finely classified ground points.
Furthermore, the position relationship between the points in the grid and the seed points includes one or more of a slope, a normal vector, a grid height difference and an absolute height, the slope is an included angle between a connecting line between the points in the grid and the seed points and a horizontal plane, the normal vector is a normal vector which uses the points in the grid as a reference, a plurality of points around the normal vector are searched and fitted to form a normal vector of one surface, the grid height difference is a difference between vertical coordinate average values of the points in the two grids, and the absolute height is a difference between vertical coordinate values of the points in the grid and the seed points.
The invention has the beneficial effects that: the method for classifying the ground points in the point cloud data screens the grids through the grid-networked point cloud data, screens the rough classification ground points according to the position relation between the points in the screened grids and the lowest points of the grids, screens the rough classification ground points, screens the fine classification ground points according to the position relation between the points in the rough classification ground point grids and the average points of the grids, and solves the problems of large ground point classification calculation amount and low precision in the prior art. Compared with the prior art, large objects are filtered in the process of roughly classifying the ground points, average points are used as seed points for finely classifying the ground points, the operation of selecting noise points or the lowest points formed by errors of original point clouds as seed points is avoided, and the overall ground point precision is improved.
Drawings
FIG. 1 is a process diagram of the method for classifying ground points in point cloud data according to the present invention.
Detailed Description
The method for classifying the ground points in the point cloud data, as shown in the attached figure 1, comprises the following steps:
s01, meshing the point cloud data;
s02, taking the lowest point in the grid as a first seed point of the grid;
s03, screening the grid according to the position relation between the points in the grid and the first seed points to obtain the screened grid;
s04, obtaining all roughly classified ground points according to the position relation between the points in the screened grid and the corresponding first seed points;
s05, performing grid-netting on all the roughly classified ground points;
s06, calculating the average point of all points in the roughly classified ground point grid, wherein the horizontal coordinate of the average point is the average value of the horizontal coordinates of all points in the grid, the vertical coordinate is the average value of the vertical coordinates of all points in the grid, and the vertical coordinate is the average value of the vertical coordinates of all points in the grid;
s07, taking the average point as a second seed point of the grid;
and S08, obtaining all the finely classified ground points according to the grids and the second seed points.
The point cloud data is subjected to gridding, then a grid with a higher relative position is filtered, then points with a higher relative position in the grid are filtered, a roughly classified ground point is obtained, then gridding is carried out on the roughly classified ground point, and a point with a higher relative position and an average point is filtered by taking the average point as a reference, so that a finely classified ground point is obtained.
Further, in step S01, the gridding the point cloud data includes:
taking the size of the maximum building as the size of the grid in the point cloud data containing the building;
and taking the first preset value as the size of the grid in the point cloud data without the building.
The point cloud data is divided by taking the size of the largest building as the size of the grid, so that the point cloud corresponding to the large building can be filtered.
Further, in step S03, the method for obtaining the filtered grid includes the following steps:
s301, calculating an average value of the position relationship between each point in each grid and the first seed point of the grid, and obtaining the grid and the corresponding average value;
s302, selecting a grid, if the grid meets the following conditions: if the average value of the grid is smaller than the average value of the adjacent grids and the difference value is smaller than the first threshold value, the grid is the screened grid;
and S303, sequentially selecting each grid to obtain the screened grids.
Further, in step S04, the method for obtaining rough classification ground points includes the following steps:
s401, selecting a screened grid, and if the position relation between a point in the grid and a first seed point corresponding to the point is smaller than a second threshold, determining that the point is a roughly classified ground point, and obtaining all roughly classified ground points in the grid;
s402, sequentially selecting all the screened grids to obtain all the roughly classified ground points.
Further, in step S05, the step of gridding all the roughly classified ground points includes: and taking the second preset value as the size of the grid.
Further, in step S08, the method for obtaining the fine classification ground points includes the following steps:
s801, selecting a grid with the average point as a second seed point, wherein if the position relation between the point in the grid and the second seed point corresponding to the point is smaller than a third threshold value, the point is a fine classification ground point, and all fine classification ground points in the grid are obtained;
s802, sequentially selecting all grids taking the average points as second seed points to obtain all the finely classified ground points.
Furthermore, the position relationship between the points in the grid and the seed points includes one or more of a slope, a normal vector, a grid height difference and an absolute height, the slope is an included angle between a connecting line between the points in the grid and the seed points and a horizontal plane, the normal vector is a normal vector which uses the points in the grid as a reference, a plurality of points around the normal vector are searched and fitted to form a normal vector of one surface, the grid height difference is a difference between vertical coordinate average values of the points in the two grids, and the absolute height is a difference between vertical coordinate values of the points in the grid and the seed points.
Example (b):
one embodiment of the method for classifying the ground points in the point cloud data, which takes the gradient as a screening condition, comprises the following steps:
s01, meshing the point cloud data:
if the point cloud data contains buildings, dividing by taking the size of the largest building as the size of the grid;
if the point cloud data does not contain buildings, dividing by taking a first preset value as the size of the grid; the first preset value is larger than the size of the non-ground-slice object;
this process can filter large buildings in the point cloud.
S02, taking the lowest point in the grid as a first seed point of the grid;
s03, calculating the gradient value of the vector from each point in the grid to the first seed point of the grid on the horizontal plane, and calculating the average value; obtaining the average gradient of all grids, selecting one grid, and if the grid meets the following conditions: if the average value of the grid is smaller than the average value of the adjacent grids and the difference value is smaller than the first threshold value, the grid is the screened grid;
the process is to screen grids and filter out grids with higher relative positions.
S04, selecting a screened grid, if the gradient value of the vector from the point in the grid to the first seed point of the grid on the horizontal plane is less than a second threshold value, the point is a rough classification ground point, and selecting all screened grids in sequence to obtain all rough classification ground points;
this process filters out points in the grid that are higher in position relative to the first seed point.
S05, performing grid meshing on all the roughly classified ground points, wherein the size of a grid is a second preset value;
and S06, calculating the average point of all points in the roughly classified ground point grid, wherein the horizontal coordinate of the average point is the average value of the horizontal coordinates of all points in the grid, the vertical coordinate is the average value of the vertical coordinates of all points in the grid, and the vertical coordinate is the average value of the vertical coordinates of all points in the grid.
S07, taking the average point as a second seed point of the grid;
and S08, selecting a grid with the average point as the second seed point, if the gradient value of the vector from the point in the grid to the second seed point of the grid on the horizontal plane is less than a third threshold value, the point is a fine classification ground point, obtaining all fine classification ground points in the grid, and sequentially selecting all grids with the average point as the second seed point to obtain all fine classification ground points.
Another embodiment of the method for classifying ground points in point cloud data of the present invention uses normal vectors as screening conditions, and comprises the following steps:
s01, meshing the point cloud data:
if the point cloud data contains buildings, dividing by taking the size of the largest building as the size of the grid;
if the point cloud data does not contain buildings, dividing by taking a first preset value as the size of the grid;
s02, taking the lowest point in the grid as a first seed point of the grid;
s03, searching a plurality of points around the grid based on a first seed point of the grid, fitting the points into a surface, calculating a normal vector corresponding to the first seed point, searching a plurality of points around the grid based on other points in the grid, fitting the points into a surface, calculating a normal vector corresponding to the points, calculating included angles between the normal vectors corresponding to all the points in the grid and the normal vectors corresponding to the first seed point, calculating an average included angle of the normal vectors, obtaining the average included angles of the normal vectors of all the grids, selecting one grid, and if the average included angles of the normal vectors of the grid are all smaller than the average included angles of the normal vectors of adjacent grids by a first threshold value, taking the grid as the screened grid;
s04, selecting a screened grid, if the included angle between the normal vector corresponding to the grid and the normal vector corresponding to the first seed point of the grid is smaller than a second threshold, the point is a rough classification ground point, and selecting all screened grids in sequence to obtain all rough classification ground points;
s05, performing grid meshing on all the roughly classified ground points, wherein the size of a grid is a second preset value;
and S06, calculating the average point of all points in the roughly classified ground point grid, wherein the horizontal coordinate of the average point is the average value of the horizontal coordinates of all points in the grid, the vertical coordinate is the average value of the vertical coordinates of all points in the grid, and the vertical coordinate is the average value of the vertical coordinates of all points in the grid.
S07, taking the average point as a second seed point of the grid;
and S08, selecting a grid with the average point as the second seed point, if the included angle between the normal vector corresponding to the point in the grid and the normal vector corresponding to the second seed point of the grid is smaller than a third threshold, the point is the fine classification ground point, obtaining all the fine classification ground points in the grid, and sequentially selecting all the grids with the average point as the second seed point to obtain all the fine classification ground points.
In the two embodiments of the present invention, the first preset value, the second preset value, the first threshold, the second threshold, and the third threshold are different in the values set in different point cloud data, and the fine classification ground points of the point cloud data obtained can be adjusted by debugging the first preset value, the second preset value, the first threshold, the second threshold, and the third threshold, so that the obtained fine classification ground points are more accurate.
The embodiment of the method for classifying the ground points in the cloud data can also take one or more combinations of gradient, normal vector, grid height difference and absolute height as a screening condition. The height difference of the grids is used as a screening condition, the average value of the vertical coordinates of the points in the adjacent grids is compared when the grids are screened, and the absolute value of the difference between the points and the vertical coordinates in the grids is used as the screening condition for the classification of the ground points. The absolute height is used as a screening condition, the average value of the difference between the vertical coordinates of the points in the adjacent grids and the seed points of the grids is compared when the grids are screened, and the absolute value of the difference between the points and the vertical coordinates in the grids is used as a screening condition for the classification of the ground points.
Claims (6)
1. The method for classifying the ground points in the point cloud data is characterized by comprising the following steps of:
s01, meshing the point cloud data;
s02, taking the lowest point in the grid as a first seed point of the grid;
s03, screening the grid according to the position relation between the points in the grid and the first seed points to obtain the screened grid; the method for obtaining the screened grid comprises the following steps:
s301, calculating an average value of the position relationship between each point in each grid and the first seed point of the grid, and obtaining the grid and the corresponding average value;
s302, selecting a grid, if the grid meets the following conditions: if the average value of the grid is smaller than the average value of the adjacent grids and the difference value is smaller than the first threshold value, the grid is the screened grid;
s303, sequentially selecting each grid to obtain the screened grids;
s04, obtaining all roughly classified ground points according to the position relation between the points in the screened grid and the corresponding first seed points;
s05, performing grid-netting on all the roughly classified ground points;
s06, calculating the average point of all points in the roughly classified ground point grid, wherein the abscissa of the average point is the average value of the abscissas of all the points in the grid, the ordinate is the average value of the ordinate of all the points in the grid, and the ordinate is the average value of the ordinate of all the points in the grid;
s07, taking the average point as a second seed point of the grid;
and S08, obtaining all the finely classified ground points according to the grids and the second seed points.
2. The method for classifying ground points in point cloud data according to claim 1, wherein in step S01, the gridding of the point cloud data comprises:
taking the size of the maximum building as the size of the grid in the point cloud data containing the building;
and taking the first preset value as the size of the grid in the point cloud data without the building.
3. The method for classifying ground points in point cloud data according to claim 1, wherein in step S04, the method for obtaining rough classification ground points comprises the following steps:
s401, selecting a screened grid, and if the position relation between a point in the grid and a first seed point corresponding to the point is smaller than a second threshold, determining that the point is a roughly classified ground point, and obtaining all roughly classified ground points in the grid;
s402, sequentially selecting all the screened grids to obtain all the roughly classified ground points.
4. The method for classifying ground points in point cloud data according to claim 1, wherein the step S05 of gridding all the roughly classified ground points comprises: and taking the second preset value as the size of the grid.
5. The method for classifying ground points in point cloud data according to claim 1, wherein in step S08, the method for obtaining the finely classified ground points comprises the following steps:
s801, selecting a grid with the average point as a second seed point, wherein if the position relation between the point in the grid and the second seed point corresponding to the point is smaller than a third threshold value, the point is a fine classification ground point, and all fine classification ground points in the grid are obtained;
s802, sequentially selecting all grids taking the average points as second seed points to obtain all the finely classified ground points.
6. The method of any one of claims 1 to 5, wherein the position relationship between the points in the mesh and the seed points comprises one or more of a slope, a normal vector, a height difference of the mesh, and an absolute height, wherein the slope is an angle between a connecting line between the points in the mesh and the seed points and a horizontal plane, the normal vector is a normal vector obtained by searching a plurality of points around the point in the mesh based on the point in the mesh and fitting the points to a plane, the height difference of the mesh is a difference between the average values of vertical coordinates of the points in the two meshes, and the absolute height is a difference between the vertical coordinates of the points in the mesh and the vertical coordinates of the seed points.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110925166.4A CN113627353B (en) | 2021-08-12 | 2021-08-12 | Method for classifying ground points in point cloud data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110925166.4A CN113627353B (en) | 2021-08-12 | 2021-08-12 | Method for classifying ground points in point cloud data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113627353A CN113627353A (en) | 2021-11-09 |
| CN113627353B true CN113627353B (en) | 2022-07-01 |
Family
ID=78384942
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110925166.4A Active CN113627353B (en) | 2021-08-12 | 2021-08-12 | Method for classifying ground points in point cloud data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113627353B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114119998B (en) * | 2021-12-01 | 2023-04-18 | 成都理工大学 | Vehicle-mounted point cloud ground point extraction method and storage medium |
| CN117496464B (en) * | 2023-10-23 | 2024-05-24 | 广东灵锶智能科技有限公司 | Ground detection method and device for foot robot |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103679655A (en) * | 2013-12-02 | 2014-03-26 | 河海大学 | LiDAR point cloud filter method based on gradient and area growth |
| CN103745441A (en) * | 2014-01-08 | 2014-04-23 | 河海大学 | Method of filtering airborne LiDAR (Light Detection and Ranging) point cloud |
| CN106157309A (en) * | 2016-07-04 | 2016-11-23 | 南京大学 | A kind of airborne LiDAR ground point cloud filtering method based on virtual Seed Points |
| US20180081035A1 (en) * | 2016-09-22 | 2018-03-22 | Beijing Greenvalley Technology Co., Ltd. | Method and device for filtering point cloud data |
-
2021
- 2021-08-12 CN CN202110925166.4A patent/CN113627353B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103679655A (en) * | 2013-12-02 | 2014-03-26 | 河海大学 | LiDAR point cloud filter method based on gradient and area growth |
| CN103745441A (en) * | 2014-01-08 | 2014-04-23 | 河海大学 | Method of filtering airborne LiDAR (Light Detection and Ranging) point cloud |
| CN106157309A (en) * | 2016-07-04 | 2016-11-23 | 南京大学 | A kind of airborne LiDAR ground point cloud filtering method based on virtual Seed Points |
| US20180081035A1 (en) * | 2016-09-22 | 2018-03-22 | Beijing Greenvalley Technology Co., Ltd. | Method and device for filtering point cloud data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113627353A (en) | 2021-11-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113627353B (en) | Method for classifying ground points in point cloud data | |
| KR101740259B1 (en) | Auto segmentation method of 3d point clouds | |
| JP4545219B1 (en) | Analysis method of topographic change using topographic image and program thereof | |
| CN108648271B (en) | Interpolation method for generating complex terrain grid model based on GIS data | |
| US9377546B2 (en) | Automatic extraction and characterization of fault and fracture populations | |
| CN111598770A (en) | Object detection method and device based on three-dimensional data and two-dimensional image | |
| CN105388513B (en) | The method for building up and device of earthquake-capturing observation system | |
| CN110363299B (en) | Spatial case reasoning method for outcrop rock stratum layering | |
| CN110807781A (en) | Point cloud simplification method capable of retaining details and boundary features | |
| CN114387506A (en) | Transmission tower monitoring method and device, computer equipment and storage medium | |
| CN105700016A (en) | Organic bank reservoir earthquake prediction method | |
| US10885662B2 (en) | Depth-map-based ground detection method and apparatus | |
| CN114511546A (en) | Laser point cloud forest breast diameter obtaining method based on DBSCAN clustering and four quadrants | |
| CN107941194B (en) | A method of obtaining and calculate engineering rock mass representativeness RQD value | |
| CN110114802B (en) | Topographic information processing device, topographic information processing method, and storage medium | |
| CN113643444A (en) | Gradient compensation-based progressive encryption triangular dot cloud filtering method | |
| Ismail et al. | DEM derived from photogrammetric generated DSM using morphological filter | |
| CN113139982B (en) | Automatic segmentation method for indoor room point cloud | |
| CN111080536A (en) | Self-adaptive filtering method for airborne laser radar point cloud | |
| Yilmaz et al. | Ground filtering of a uav-based point cloud with the cloth simulation filtering algorithm | |
| CN113655498B (en) | Method and system for extracting cone barrel information in racetrack based on laser radar | |
| CN113686600A (en) | Performance identification device for rotary cultivator and ditcher | |
| CN108986217B (en) | Multipoint geostatistical modeling method based on pattern vector distance | |
| CN114820400B (en) | Airborne LiDAR point cloud ground point filtering method | |
| CN113362340A (en) | Dynamic space sphere searching point cloud K neighborhood method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |