Obtenção da altura de louro-pardo a partir de imagens digitais em sistema silvipastoril

Abstract

The use of drones has been growing in several areas of research, both due to their agility and practicality in relation to traditional methods of aerial photogrammetry. With the insertion of these new technologies within the forestry sector, helping in the planning and management of resources, the possibility arises of testing new techniques either to improve or to present new methodologies in different stages of the forest production chain. The objective of this work is to find out how accurate are the height measurements obtained from the analysis of high resolution digital images obtained with an RPA within a forest plantation located within the UTFPR - Câmpus Dois Vizinhos, in an experimental unit that uses the silvopastoral with the laurel-pardo forest species (Cordia trichotoma(Vell.) Arrab. ex Steud) and goat and sheep farming. A flight with RPA was carried out on site and high resolution digital images were collected, later processed within the Open Drone Map software, where through the generation of apoint cloud it was possible to obtain products such as an orthomosaic of the area, as well as the digital elevation and the digital surface model. Then, the orthomosaic processing and digital models of elevation and surface generated by the software were carried out within a GIS environment, where using the available tools, the altitude values of the digital models were extracted, and these values were used to calculate the height of each tree at planting. These values were then compared with real measurements, measured during a thinning carried out in the same month of the flight, in the study area. As well, with measurements taken using a hypsometer during the last inventory before thinning, again, for comparison purposes. With all these data, the R Studio software was used to compare both graphically and statistically the height values found in the different methodologies. After statistical analysis, it was found that the square root of the mean squared error is 1.94 meters, meaning that the heightvalues found can vary both for more and for less this value, the mean error of -0.994 meters, showing that the The drone ends up overestimating the height values by approximately 1 meter and an average relative error of -9.65, that is, the drone ends up overestimating the height of the individuals by approximately 9.6%. Therefore, it is possible to conclude that for dense plantations, the methodology of obtaining height values from high resolution images captured with RPA is not valid, since the crown overlap made it impossible to identify each parrot laurel individual. Besides, there is a big difference between the real values and those found remotely. Therefore, further studies are needed in the area. In addition to a special attention when selecting the study areas, with preference for arboreal individuals that have an easily identifiable canopy and that have low canopy overlap.