- Title: Using Video Observations for Visual Navigation and Map-building
- Description: Visual navigation in robotics using cameras are becoming more popular as embedded systems are becoming more powerful. Traditional and reliable sensors are being supplemented by cameras improving accuracy and adaptability, which in turn increase the complexity of underlying software systems. At the Lund Institute of Technology, a project introducing a modular approach to system design usable with embedded systems and robotics has been started. It aims for making such systems easier to maintain and reuse. The project includes various topics, where one is navigation of a mobile robot using cameras. The aim of this thesis is to investigate and develop methods for such visual navigation. Two scenarios are investigated, the first being a ceiling-mounted camera used to find the absolute position of the mobile robot and the second being a camera mounted on the mobile robot used for building a map over the working area. For the first scenario, a geometric model is proven to be an effective approach in finding the position of a designed pattern, and also possible to extend to support different zoom levels. For the second scenario, an algorithm using correlation between images to build a map is developed. As the algorithms are targeted for real-time systems, consideration has also been taken to enable efficient implementation of the developed methods in a low-level programming language. The result is a reliable detector for the absolute position and an efficient map-building method supporting both translated and rotated images.
- Start Date: Sept. 6, 2010
- Finished Date: Dec. 17, 2010
- Supervisor: Håkan Ardö
- Supervisor: Anders Heyden
- Report (8.3 MB)
- Popular Science Report (156.3 KB)
Questions: webmaster
Last update: 2013-04-11
Centre for Mathematical Sciences, Box 118, SE-22100, Lund. Telefon: +46 46-222 00 00 (vx)