Abstract

The energy sector is nowadays experiencing a major tendency to an increase in inspection and monitoring frequency for environmental, safety and asset integrity reasons. To meet these increasingly demanding requirements the sector is naturally shifting toward remote or autonomous solutions such as robotics. Robots enable autonomous inspection and monitoring tasks, but they need localization and sensor targeting systems that enable them to perform these tasks autonomously and safely in an industrial facility. This paper addresses the problem of the localization of a robotic device (UAV) for autonomous in-plant inspections and the problem of recognizing and targeting objects to be inspected within the scene displayed by a camera on robot board, providing an example of a customized implementation of existing algorithms modified and developed to suit the industrial environment in which the drone operates. First the robot localization problem is addressed starting from existing work in literature. Afterward a comprehensive solution for visual object detection and target tracking based on combination of visual processing and machine learning techniques is proposed. Finally, tests performance of both the systems are analysed in an industrial oil and gas environment.

This content is only available via PDF.
You can access this article if you purchase or spend a download.