Qualidade dos produtos de aerofotogrametria por drone para uma floresta madura na Amazônia central

Abstract

Images from Remotely Piloted Aircraft (RPA) have been widely used in the forestry sector and in academia, however, there are still issues related to accuracy, quality and potential applications of the information acquired. Monitoring the forest from an aerial perspective by aerophotogrammetric remote sensing is one way to spatially analyze the structure and the effects of disturbances on forest dynamics, with the possibility of enhancing the results when related to field data from the forest inventory (FI). Several authors have found that to achieve the quality of these products requires rigor in the acquisition, especially because we are producing mathematical models of a rough, irregular and extremely dynamic surface by means of computer vision algorithms, such as Structure from Motion (SfM). Furthermore, there is no consensus on acquisition protocols and parameterization for specific questions related to forest types and environmental conditions. The goal of this project is to evaluate the geometric and radiometric quality of aerophotogrammetry products based on metrics extracted from the forest structure, obtained by different acquisition methods with ARP. Here the results suggest that different combinations altitude, overlap and illumination can influence the quality of PRA products. For most methods there were no major differences in mean canopy area and perimeter. However, quality can be achieved by combining diffuse light illumination (cloudy sky), high overlap, and proximity to the target.