Možnosti filtrácie vegetácie z bodových mračien získaných UAV

Abstract

The aim of the thesis is to evaluate possibilities of vegetation filtering from point clouds derived from RGB images acquired by unmanned aerial vehicle using the technique Structure From Motion. Three selected software packages are tested and compared. The introduction of the thesis describes the possibilities of creating digital terrain models using unmanned aerial vehicles and briefly explains the procedures of selected filtering algorithms. Thesis also describes the characteristics of test sites and process of point clouds filtering in individual software. Finally, the software packages are evaluated in terms of filtration effectivity, user-friendliness and height accuracy validated on a set of control points measured by electronic tachymetry. Moreover, digital terrain models derived from the resulting filtered point cloud are compared to each other. Study was realized in three software packages: Agisoft PhotoScan Professional, LAStools and 3DReshaper. Each software provided similar results, however during procesing each of them revealed its advantages and disadvantages, which are discussed in detail in individual chapters.