Catadioptric vision for robotic applications

Abstract

Vision is a very important sense and is essential to humans and animals. It permits to perceive, manipulate, model the environment and also navigate into it. The vision ability is mimicked for robots using cameras. However conventional cameras suffer from a small field of view, which might lead to important limitations and difficulties. This PhD thesis is rather dedicated to omnidirectional vision systems, especially catadioptric cameras, that permit to overcome these difficulties. An omnidirectional camera can see in every direction or at least has a very wide field of view. A catadioptric camera is a special kind of omnidirectional system: it is composed of a camera and a convex mirror, which permits to greatly enlarge the field of view and thus acquire much more information from the environment. The objective of this thesis is to study the role of omnidirectional vision for robotic applications such as low-level feature extraction (lines, skyline, object) but also motion estimation and 3D reconstruction in urban environment, while facing the fundamental issues of omnidirectional vision. Our work has also been motivated by the "mega-city modeling" project, performed in collaboration with Ikeuchi`s laboratory at the University of Tokyo, Japan. Several algorithms have been proposed in the literature to estimate the motion of a catadioptric camera. Combined with existing quasi-dense matching technique and texture mapping methods, some existing methods can obtain impressive results for urban 3D reconstruction. It is important to note that most of these methods are simply based on textbook materials and do not really take into account the inherent aspects of omnidirectional vision. On the contrary, our work aims to face some fundamental issues associated to omnidirectional vision, especially image distortion and repetitive textures/interest points. These issues are often related to each others. For example, because of distortions, it becomes difficult to match cor...