Využití bezpilotních leteckých prostředků v inventarizaci lesních porostů

Abstract

The thesis presents a single tree segmentation method applied to 3D point cloud data acquired with a LiDAR scanner mounted on an unmanned aerial vehicle. The method is applied directly to the point cloud and is based on graph theory and clustering. The method uses only the geometric properties (position) of the points to divide them into clusters. These are then connected by edges to form a 3D graph to which the graph partition function is applied. Final adjustments are then made to obtain the tree segments. To test the method, different data sets were acquired at three test sites in the Czech Republic. The sites are covered by economic forest and were imaged in the vegetation season with different flight parameters to evaluate their influence on the segmentation results. The results were compared with a reference dataset taken during field reconnaissance within the framework of the National Forest Inventory and a reference dataset taken based on manual tree detection on an orthophoto image. In the final part of the thesis, the influence of data collection parameters on segmentation was evaluated, and the result of the presented method was compared with a commercial solution.