Estudo de métodos para detecção de torres e posicionamento de multirrotores: aplicação de VANT em inspeções de sistemas de energia

Abstract

Inspections of power towers must be carried out routinely, but it is still often carried out manually, resulting in risks for the technicians involved in the process. Advances in the areas of Unmanned Aerial Vehicles (UAV) currently allow this task to be carried out with the use of these aircraft, with the use of on-board sensors and with the aircraft being manually piloted. While it is theoretically possible to automate this process, there are some barriers that prevent autonomous operation from being fully reliable, including correct error detection and collision avoidance, and the need for redundant precision positioning in the event of sensor failures. The goal of this work is to study tools and methods focused on solving the pylon detection and positioning issues. For this, the contribution of this work is to evaluate low-cost sensors that can be easily embedded in aircraft, and two methods that can complement the positioning configuration of aircraft in the context of eletric energy tower inspection: a positioning method using an AR-Tag as an external reference, and a detection and orientation method in relation to towers using planar LiDAR sensors and machine learning. The results show that the detection of results found in the energy tower environment is possible with the low-cost sensors tested. The proposed positioning methods, although still in need of improvements, also generated promising results, such as an accuracy with the use of tags superior to a common GPS, and a high rate of correct answers in the identification of towers with only the LiDAR sample. These methods can prove useful to carry out the inspection in the safest and most efficient way.