The Retrieval Of The Geometrical Structure Parameters Of Single Trees Based On The Terrestrial Lidar Scanning Method

Accurate investigation of forest resources is critical for the management of forest resources and planning of forestry policy. Traditional direct methods of forest inventory have the disadvantages of labor-intensive, time-consuming, low efficiency; they can’t match the application demands of the modern precision forestry increasingly. As compared to the traditional direct methods, the terrestrial LiDAR scanning method has the advantages of high precision, non-destructive to trees, the penetrability of forest canopies, etc., so it was widely applied in the field of forest inventory, especially at the area of the retrieval of the geometrical structure parameters of single trees. In this paper, the point clouds of the single trees were collected based on the terrestrial scanning method first, the geometrical model of the single trees were modeled by using the simple and free-form surface methods separately, the geometrical structure parameters of the single trees were retrieved from the models of the single trees at last. The main contents of this paper are as following:(1) The collection and preprocessing of the point clouds of the single tree. A bombax tree with the height of 18m was selected as the research object. With the demand of the construction of the geometrical modeling of the single tree, a sound reasonable scheme was designed for the collection of the point clouds of the object. Registration between the points clouds of multi-stations were carried out next. Resampling was performed in order to reduce the huge number of point clouds contained in the original datasets collected and enhance the efficiency of data processing.(2) The geometrical modeling of the surface of the single tree based on the simple and free-form surface methods. As for the simple surface method, the Kd-tree data structure was chosen as the organization scheme of the point cloud datasets. The point clouds of the single tree was segmented into several sections by using the shortest path algorithm, after that, the geometric center of each sections were inferred, and a quadratic curve of circles and ellipses were fitted at each height level of the geometrical center for each section, the cylinder and cone were used as standard element for the construction of the surface of each section. But for the free-form surface scheme, the Voronoi diagram was generated based on the original point cloud data sets first, after that, the Delaunay triangulation network would be archived from the generated Voronoi diagram and original point cloud data sets, the triangles with which all the three vertexes are among from the original point clouds were regarded as the final constructed surface model for the Delaunay triangulation network in the end. Results showed that the simple surface method showed the merits of simplicity and high efficiency, but the surface model of the connection parts between the two adjacent sections was deviated from the real tree stem and point cloud data sets obviously. As compared to the simple surface method, the free-form surface scheme shows the advantage that it can model the surface of the tree stem with high quality in detail.(3) The estimation of the geometrical structure parameters of single trees based on the point cloud data sets. The geometrical structure parameters of single trees were retrieved based on the surface model constructed by using the simple and free-form surface methods. The estimate of the DBH retrieved from the constructed tree surface model was 1.1% different from the results of the direct measurements. The estimates of the tree volume calculated from the modeled surface of the single tree were-4.58%、-3.55%、2.13% different from the regression equation method.