| With the intensification of global warming trend,people generally recognize the urgency of solving climate problems.Carbon sequestration in forests has become one of the main ways to mitigate climate change in countries around the world.In the process of studying the relationship between forest ecosystems and carbon cycle,the investigation of forest resources is of key significance,and obtaining forest tree parameter can provide theoretical support and basis for the formulation of global climate change mitigation policies.How to extract accurate forest tree parameter from the complex forest structure has become a hot issue in current research.Terrestrial laser scanning can collect detailed forest interior information and has significant advantages in extracting the interior parameter of forest.Therefore,this paper uses terrestrial laser scanning point cloud data to extract tree parameter.Three important steps are included in the parameter extraction process: point cloud filtering,single tree extraction and tree parameter estimation.The research work of this paper consists of the following three parts.1 A multi-scale filtering method based on terrestrial LiDAR point clouds is proposed.Point cloud filtering is the process of extracting ground points from point cloud data.In complex forest scenes,the existing filtering algorithms generally have the problems of difficult to confirm the filtering threshold,unpreserved terrain details,and unstable filtering accuracy.To solve such problems,the filtering method in this paper uses the least squares method to fit the surface with the ground seed points under the multi-scale window,progressively constructs the digital terrain model,and gradually acquires the accurate ground point cloud data by combining morphological open operations.In this paper,three sets of forest data under uncomplexity are selected for experiments,and it can be concluded that the average total error of the three sets of sample data is 1.07%,which indicates that the method is suitable for filtering terrestrial LiDAR point cloud data in forest scenes and can obtain better filtering effect.2 A single tree segmentation method based on connectivity marker optimisation is proposed.Single tree segmentation is the extraction of single trees from point cloud data and is an important basis for subsequent tree parameter extraction.Currently,existing single tree segmentation algorithms suffer from inaccurate extraction and slow extraction speed.In order to reduce the errors of single tree extraction,this paper proposes to use the connectivity of local maxima under small windows to determine the marker points,and use the watershed algorithm to achieve the initial extraction of single tree.Finally,the density contour property of trees is used to perform secondary extraction of under-segmented trees to obtain the final single tree segmentation results.The experimental results show that the segmentation results of the proposed method are better than the Mean shift single tree segmentation method and the traditional marker-based watershed single tree segmentation method in three groups of different complexity sample areas.3 The method of estimation of forest tree parameter is studied.Forest tree parameter are important measurements in forestry surveys and are important indicators for assessing forest growth.In this paper,four parameter of tree height,diameter at breast height,location and crown width are estimated to provide reference for in-depth research on tree parameter estimation methods.In this paper,tree height is calculated using the top and lowest points of single tree point cloud data.The diameter at breast height is extracted by least squares circle fitting method and Hough transform method,and the center point of the diameter at breast height circle of the fitted trees is considered as the location of the trees.For the estimation method of tree crown width,this paper uses the traditional estimation method and the convex package method to find the tree crown width. |
PDF Full Text Request