Branches Modeling And Morphological Parameters Extraction Technology Based On3D Laser Scanning Technology

With the development and deepening of forest visualization and forestry mappingtechnology, the process of building Digital Forestryfor our country continues to move forward.As a new mapping technology, terrestrial3D laser scanning does not only expand the technicalmeans of achieving virtual scene and generating tree growth model from the real world, butalso can collect three-dimensional data of trees truly, directly and with highprecision in digitalformat. This provides an easy way for surveying forest resources, studying stand structure aswell as reconstructing tree3D model. In this study, reconstruction of single tree3D models wasconducted. Based on point cloud data from terrestrial3D laser scanning, this study builtcontour modeland Delaunay triangulated irregular network (TIN) model of tree stems andbranches and achieved the automatical and manual extraction of tree morphology and structureparameters.Point cloud data collected by3D laser scanner for tree stems and branches were dividedinto different layers along the vertical direction and the hierarchical model of the tree was built.Using convex hull algorithm, the point cloud data along the contours of different heightswerethen extracted, leading to contours of tree stem and branches. The Delaunay TINs betweenadjacent contours was developed using TIN growth algorithm. Merging the Delaunay TINs atdifferent heights of the tree led to the tree Delaunay TIN model. This method was validatedusing an example. Compared with other methods, this approach did not only require less databut also led to better results.According to the tree and branch Delaunay TIN model, tree morphology parameters wereautomatically extracted. Using interactive ray collision detection technology, theparameterswere manually obtained.As an example, the tree morphology parameters of afour-year old Liriodendron in Huangfengqiao state-owned forest farm were extracted andcompared with the field observations. The obtained error of the model parameterswas less than5%, which meetsthe requirement of tree measurement in forestry. Moreover, a system for extracting the parameters of stem and branches was developedusing C#languageand theobject-oriented approachbased on DirectX platform.