Method for acquiring topographic information under vegetation shielding of artificial forest area
By using oblique photography from drones and an improved clustering algorithm to filter tree vertex coordinates, and combining this with DEM data to fit topographic maps, the problem of obtaining topographic information under the obstruction of artificial forests was solved, achieving an efficient and low-cost method.
Patent Information
- Application Number
- CN202510831202.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-20
- Publication Date
- 2025-11-11
AI Technical Summary
Existing technologies struggle to efficiently and cost-effectively acquire topographic information in plantation areas where vegetation obscures the terrain. Traditional methods are inefficient, airborne lidar is expensive, and drone photogrammetry cannot obtain surface information obscured by vegetation.
3D point cloud data was acquired through oblique photography by UAVs. The tree vertex coordinates were selected by combining improved Euclidean clustering and convex hull algorithms. The topographic map was then fitted with DEM elevation data to reduce the density of the point cloud data and to crop it, thereby selecting the ground point coordinates.
It enables low-cost and efficient acquisition of terrain information under the cover of artificial forests, improves computing efficiency and accuracy, and can generate complete topographic maps.
Smart Images
Figure CN120931844A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of point cloud data processing, specifically relating to a method for obtaining terrain information under vegetation cover in artificial forest areas. Background Technology
[0002] Acquiring topographic information is an important research direction in the field of surveying and mapping, and a key parameter in engineering geology, widely used in exploration and design, disaster monitoring, and other fields. Currently, methods for acquiring topographic information can be divided into three categories: traditional surveying, radar remote sensing, and optical remote sensing. Traditional surveying was an early method for landslide monitoring, mainly relying on manual investigation and screening. The focus of monitoring was slope deformation, and common deformation detection methods included leveling, GPS surveying, and deep displacement gauges. However, these methods have drawbacks: the observation methods are all point-based, resulting in high costs and low efficiency, making them unsuitable for large-scale detection. Radar remote sensing, on the other hand, uses imaging sensors to transmit electromagnetic waves to the ground, then receives the electromagnetic waves reflected back from the ground, records and stores the received electromagnetic waves, and generates SAR images. Airborne radar laser scanning can provide high-resolution, high-precision topographic images. Simultaneously, through multiple-echo technology, it penetrates ground vegetation, and using filtering algorithms, it effectively removes surface vegetation to obtain true ground elevation data, providing an important means for identifying high-altitude, concealed geological hazards. Satellite optical remote sensing technology, due to its high timeliness, strong macroscopic perspective, and rich information, has become an important technical means for the investigation, analysis, and assessment of major natural disasters. With the rapid advancement of UAV technology, UAVs can be used to conduct high-precision vertical aerial and oblique photogrammetry, and quickly generate digital topographic maps, digital orthophoto maps, digital surface models, and digital ground models of the survey area. This allows for the discovery and identification of geological hazard risks, and also enables the calculation of vertical surface displacement, volume changes, and profiles before and after changes.
[0003] Current methods for acquiring topographic data all need to address the problem of vegetation interference. For hilly areas covered by large areas of planted forest, manual surveying is too inefficient and unsuitable for topographic measurement. Airborne LiDAR technology requires filtering to remove surface vegetation; furthermore, airborne LiDAR is expensive and not suitable for large-scale applications. Unmanned aerial vehicle (UAV) photogrammetry cannot obtain information about the surface obscured by vegetation. In conclusion, there is currently no cost-effective method for acquiring topographic information in forested areas. Summary of the Invention
[0004] Purpose of the invention: To address the above-mentioned shortcomings, this invention provides a method for obtaining topographic information under vegetation cover in artificial forest areas, thereby achieving accurate acquisition of topographic information under forest cover.
[0005] Technical solution: The present invention provides a method for obtaining topographic information in plantation forest areas under vegetation cover, comprising the following steps:
[0006] (1) Obtain and process the three-dimensional point cloud data of the study area through UAV oblique photography and modeling;
[0007] (2) An improved Euclidean clustering algorithm is used to cluster point cloud data, and a kd-tree algorithm is used to build a kd-tree of point cloud coordinate values for neighbor search.
[0008] (3) The tree vertex coordinates are selected using an improved convex hull algorithm;
[0009] (4) Subtract the tree height h from the z value of all recorded points to obtain the coordinates of the ground points;
[0010] (5) Convert the DEM elevation data into point coordinates and place them in the same coordinate system as the ground point coordinates obtained in step (3). Replace the coordinates obtained from the DEM conversion in the overlapping area with the point coordinates obtained in step (3) and perform plane fitting to obtain the topographic map of the forest-covered area.
[0011] Furthermore, the implementation process of step (1) is as follows:
[0012] Thin out the 3D point cloud data to reduce its density.
[0013] The thinned point cloud data is cropped to retain the forest area and remove the surrounding bare slopes.
[0014] The 3D point cloud data is measured, including tree height h, canopy radius, and number of canopy points. The minimum search radius r is defined based on the canopy radius, and the minimum and maximum number of points minPoints and maxPoints are defined based on the number of canopy points.
[0015] Furthermore, the improved Euclidean clustering algorithm described in step (2) is an improvement on the Euclidean clustering algorithm by using point cloud Euclidean distance, incorporating weights α for horizontal and vertical distances during distance calculation:
[0016]
[0017] Where r is the search range, D xy D represents the horizontal distance between point clouds. z This represents the vertical distance between point clouds.
[0018] Furthermore, α is 1.3.
[0019] Furthermore, in step (2), the clustering distance threshold for clustering point cloud data using the improved Euclidean clustering algorithm is set to r = 1.5m, the minimum number of point clouds per cluster is set to minPoints = 18, and the maximum number of point clouds per cluster is set to maxPoints = 40.
[0020] Furthermore, the improved convex hull algorithm described in step (3) is achieved by adding weights to the point cloud height and density in the convex hull algorithm:
[0021]
[0022] in, The weight of the point cloud height is denoted by z, and the weight of the z-axis of the point cloud is denoted by i The value and the minimum z of the point cloud within this cluster min The difference is such that the higher the point cloud height, the greater the weight of that point cloud height. The weights of the point cloud density are denoted by D(v), and the distances between the point cloud and its (k-1) nearest neighbors are denoted by D(v). i ,v ij The sum of all points is taken as the reciprocal of the average, where k is 7; the higher the point cloud density, the greater the weight of that point cloud density; W i This is the overall weight.
[0023] Furthermore, the implementation process of step (3) is as follows:
[0024] For each cluster, an improved convex hull algorithm is used to calculate a weighted point cloud convex hull model for all points within the cluster. For the vertex coordinates v of this convex hull... i Perform weighted averaging to obtain the weighted average value v of the point cloud coordinates of the convex hull. top , as the tree vertex of this cluster:
[0025]
[0026] The calculated x and y coordinates of the tree vertices are searched for values smaller than the planting spacing radius R of the artificial forest, and only the points with the maximum z value in each group are retained.
[0027] Beneficial effects: Compared with the prior art, the beneficial effects of this invention are as follows: This invention proposes a low-cost and efficient method for acquiring artificial forest terrain information; the three-dimensional point cloud model obtained by UAV oblique photogrammetry modeling is thinned and cropped to reduce data density and improve computational efficiency; neighborhood analysis and calculation are performed using an improved Euclidean clustering algorithm and an improved convex hull algorithm to filter out tree vertex coordinate information, and ground point coordinates are calculated by combining tree height and tree vertex coordinates; complete terrain information obscured by artificial forest is fitted by combining DEM elevation data. Attached Figure Description
[0028] Figure 1 This is a flowchart of the present invention;
[0029] Figure 2 The processed sparse 3D point cloud model;
[0030] Figure 3 This is a comparison diagram of the terrain surface obtained using the present invention and the original model. Detailed Implementation
[0031] The present invention will now be described in further detail with reference to the accompanying drawings:
[0032] like Figure 1 As shown, this invention proposes a method for obtaining topographic information in plantation forest areas under vegetation cover. The specific implementation process is as follows:
[0033] The study area in this embodiment is located on the S210 highway in Guangxi Zhuang Autonomous Region, and includes some exposed slopes and eucalyptus forests.
[0034] Step 1: Obtain five-dimensional imagery of the study area using oblique photography by UAV, and generate a 3D point cloud model based on the imagery, totaling 188,436,044 points. Simultaneously, generate a high-precision DEM (Digital Elevation Model). Thin the point cloud according to the octree layer count, setting the octree layer count to 7, leaving 15,013 points. After simple cropping to remove points representing roads and exposed slopes, a final balance of 7,981 points remains. The processed point cloud is shown below. Figure 2 As shown, the vast majority of the remaining points in the processed point cloud belong to the tree canopy, while a small number of points belong to the trunks and bases of trees on the outer edge of the forest.
[0035] The 3D point cloud model was measured, including tree height (h), crown diameter, and crown point cloud density. Ten eucalyptus trees from the periphery were evenly selected for height measurement (numbered 1-10), and ten treetops from the entire study area were randomly selected for crown diameter and crown point cloud density measurement (numbered 11-20). The measurement results are shown in Table 1.
[0036] Table 1. Information Measurement Results of Artificially Planted Eucalyptus Forests
[0037]
[0038] Based on the height of the eucalyptus trees, the height of the plantation is determined to be h = 15m. Based on the crown diameter, the search radius r is defined as 1.5m. Based on the crown point cloud quantity, the minimum point cloud quantity minPoints is defined as 18. Based on the crown point cloud quantity, the maximum point cloud quantity maxPoints is defined as 40.
[0039] Step 2: The improved Euclidean clustering algorithm is used to cluster the point cloud data, and the kd-tree algorithm is used to build a kd-tree of the point cloud coordinate values for subsequent neighbor search.
[0040] To adapt to the characteristics of treetops in plantations, this invention improves upon the Euclidean clustering algorithm by using point cloud Euclidean distance. Weights for horizontal and vertical distances are added during distance calculation, as shown in the following formula:
[0041]
[0042] Among them, D xy D represents the horizontal distance between point clouds. z Let be the vertical distance between point clouds. The weight of the horizontal distance, α, is 1.3, and the weight of the vertical distance, (2-α), is 0.7. The value of α should be adjusted according to the specific situation. This method reduces the weight of the vertical distance, so that the clustering algorithm pays more attention to the vertical distance when searching for neighboring points with r as the search range, which greatly reduces the original algorithm's mistaken clustering of branches and leaves of adjacent trees.
[0043] In this specific implementation, the clustering distance threshold is set to r = 1.5m, the minimum number of points in each cluster is set to minPoints = 18, the maximum number of points in each cluster is set to maxPoints = 40, and the kdtree method is selected for searching neighboring points.
[0044] Step 3: Based on the convex hull algorithm, weights were assigned to the height and density of the point cloud to filter out the coordinates of the eucalyptus tree vertices.
[0045]
[0046] Among them, the weight of point cloud height For the z-axis of the point cloud i The value and the minimum z of the point cloud within this cluster min The difference in point cloud density weights is as follows: the higher the point cloud height, the greater the weight of that point cloud height. Let D(v) be the distance between the point cloud and its (k-1) neighboring points. i ,v ij The sum of the points and the reciprocal of their average values are taken, where k = 7. Higher point cloud density results in a higher density weight for that point cloud. After calculating the height and density weights of all points in the cluster, the height weights of all point clouds are... and density weight Multiply to determine the overall weight W for each point cloud. iThis method reduces the impact of noise and local anomalies within clusters on the results.
[0047] For each cluster, an improved convex hull algorithm is used to calculate a weighted point cloud convex hull model for all points within the cluster. For the vertex coordinates v of this convex hull... i Perform weighted averaging to obtain the weighted average value v of the point cloud coordinates of the convex hull. top This is taken as the eucalyptus vertex of the cluster:
[0048]
[0049] The x and y coordinates of the eucalyptus tree vertices obtained above were searched using a radius R smaller than the planting spacing radius R of this plantation (2m in this case), retaining only the point with the maximum z value in each group. After calculation, a total of 76 vertices were obtained.
[0050] Step 4: Subtract 15 from all the z-coordinate values of the eucalyptus tree vertices to obtain the z-coordinate values of the ground.
[0051] Step 5: Convert the DEM elevation data into point coordinates. The DEM resolution is 1m, and the generated point coordinates are spaced 1m apart in both the x and y directions. Place them in the same coordinate system as the ground point coordinates obtained in Step 3. Replace the coordinates obtained from the DEM conversion in the overlapping areas with the point coordinates obtained in Step 3, and perform plane fitting to obtain the topographic map of the eucalyptus forest-covered area. Then, generate the topographic information based on the point coordinate information.
[0052] Figure 3 The figure shows a comparison between the fitted terrain surface and the original model. It can be seen that the surface successfully restored the terrain information that was obstructed by eucalyptus forests in the original model, and it has good accuracy. The method of this invention can obtain terrain information under artificial forest cover in a low-cost and efficient manner, and can then be reconstructed into a 3D model for further analysis. This provides effective information for natural resource statistics, surveying of surrounding projects, and investigation of geological hazards.
[0053] The above content merely illustrates the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made based on this technical solution fall within the scope of protection of the technical concept proposed in this invention and are included in the claims of this invention.
Claims
1. A method for obtaining topographic information under vegetation cover in artificial forest areas, characterized in that, Includes the following steps: (1) Obtain and process the three-dimensional point cloud data of the study area through UAV oblique photography and modeling; (2) An improved Euclidean clustering algorithm is used to cluster point cloud data, and a kd-tree algorithm is used to build a kd-tree of point cloud coordinate values for neighbor search. (3) The tree vertex coordinates are selected using an improved convex hull algorithm; (4) Subtract the tree height h from the z value of all recorded points to obtain the coordinates of the ground points; (5) Convert the DEM elevation data into point coordinates and place them in the same coordinate system as the ground point coordinates obtained in step (3). Replace the coordinates obtained from the DEM conversion in the overlapping area with the point coordinates obtained in step (3) and perform plane fitting to obtain the topographic map of the forest-covered area.
2. The method for obtaining topographic information under vegetation obstruction in an artificial forest area according to claim 1, characterized in that, The implementation process of step (1) is as follows: Thin out the 3D point cloud data to reduce its density. The thinned point cloud data is cropped to retain the forest area and remove the surrounding bare slopes. The 3D point cloud data is measured, including tree height h, canopy radius, and number of canopy points. The minimum search radius r is defined based on the canopy radius, and the minimum and maximum number of points minPoints and maxPoints are defined based on the number of canopy points.
3. The method for obtaining topographic information under vegetation obstruction in an artificial forest area according to claim 1, characterized in that, The improved Euclidean clustering algorithm described in step (2) is based on the Euclidean clustering algorithm, but uses point cloud Euclidean distance for improvement. Weights α for horizontal and vertical distances are added during distance calculation. Where r is the search range, D xy D represents the horizontal distance between point clouds. z This represents the vertical distance between point clouds.
4. The method for obtaining topographic information under vegetation obstruction in an artificial forest area according to claim 3, characterized in that, The value of α is 1.
3.
5. The method for obtaining topographic information under vegetation obstruction in an artificial forest area according to claim 1, characterized in that, In step (2), the improved Euclidean clustering algorithm is used to cluster the point cloud data. The clustering distance threshold is set to r = 1.5m, the minimum number of point clouds in each cluster is set to minPoints = 18, and the maximum number of point clouds in each cluster is set to maxPoints = 40.
6. The method for obtaining topographic information under vegetation obstruction in an artificial forest area according to claim 1, characterized in that, The improved convex hull algorithm described in step (3) is to add weights to the point cloud height and density in the convex hull algorithm, assigning them as follows: in, The weight of the point cloud height is denoted by z, and the weight of the z-axis of the point cloud is denoted by i The value and the minimum z of the point cloud within this cluster min The difference is such that the higher the point cloud height, the greater the weight of that point cloud height. The weights of the point cloud density are denoted by D(v), and the distances between the point cloud and its (k-1) nearest neighbors are denoted by D(v). i ,v ij The reciprocal of the average of the sums of the points (W) indicates that the higher the point cloud density, the greater the weight of that point cloud density. i This is the overall weight.
7. The method for obtaining topographic information under vegetation obstruction in an artificial forest area according to claim 1, characterized in that, The implementation process of step (3) is as follows: For each cluster, an improved convex hull algorithm is used to calculate a weighted point cloud convex hull model for all points within the cluster. For the vertex coordinates v of this convex hull... i Perform weighted averaging to obtain the weighted average value v of the point cloud coordinates of the convex hull. top , as the tree vertex of this cluster: The calculated x and y coordinates of the tree vertices are searched for values smaller than the planting spacing radius R of the artificial forest, and only the points with the maximum z value in each group are retained.
8. The method for obtaining topographic information under vegetation obstruction in an artificial forest area according to claim 1, characterized in that, The value of k is 7.