Study on passive and active computer vision paradigms related to UAVs movements for electrical pylons analysis

Student thesis: Master typesMaster in Computer Science Professional focus in Data Science

Abstract

This master thesis aims to make a first step in the research of using simulation for UAVs in the domain for computer vision applications. We will compare passive vision and active vision paradigms. Passive vision is a system where the computer takes a decision depending on the actual image it receives. Active vision, on the other hand, manipulates the viewpoint to investigate the environment and retrieve better pieces of information. We were able to achieve a fully running passive vision U-Net model. That model can successfully navigate around a pylon following a path that covers the whole pylon. The active approach was more complicated. It was only possible to do a reflection plus some unsuccessful tests. A time comparison was made. We have also compared the complexity of modifying in case of enterprise use. U-Net showed to be easier to use and change compared to the active vision paradigm by a far hand. Finally, we open the possibilities of future researches in that still yet to exploit domain.
Date of Award24 Jun 2020
Original languageEnglish
Awarding Institution
  • University of Namur
SupervisorElio Tuci (Supervisor)

Keywords

  • Computer vision
  • Active vision
  • Passive vision
  • Autonomous UAV
  • Intelligent system
  • Artificial intelligence
  • U-Net
  • CTRNN

Cite this

'