Percorrer por autor "Oliveira, Daniel dos Santos"
A mostrar 1 - 1 de 1
Resultados por página
Opções de ordenação
- Inspecção Visual de Isoladores Eléctricos -Abordagem baseada em Deep LearningPublication . Oliveira, Daniel dos Santos; Dias, André Miguel PinheiroTo supply the electrical population’s demand is necessary to have a good quality power distribution systems. Electrical asset inspection, like electrical towers, dam or power line is a high risk and expensive task. Nowadays it is done with traditional methods like using a helicopter equipped with several sensors or with specialised human labour. In the last years, the Unmanned Aerial Vehicle (UAV) exponential growth (most common called drones) make them very accessible for different applications. They are cheaper and easy to adapt. Adopting this technology will be in the future the next step on electrical asset inspection. It will provide a better service (safer, faster and cheaper), particularly in power line distribution. This thesis brings forward an alternative to traditional methods using a UAV for images processing during the insulator visual inspection. The developed work implement real-time insulators visual detection using na Artificial Neural Network (ANN), You Only Look Once (YOLO) in this case, on medium and high voltage power lines. YOLO was trained with different types and sizes of insulators. Isn’t always possible to see what the UAV is recording so it has a gimbal system which controls the camera orientation/position. It will centre the insulator on the image and this way getting a better view of it. All the training and tests were performed on board Jetson TX2.