Localization of Sensors and Objects in Distributed Omnidirectional Vision

Recent progress of multimedia and computer graphics is developing practical application systems based on simple computer vision techniques. Especially, the practical approach recently focused on is to use multiple vision sensors with simple visual processing. The objective of this research is to introduce a distributed omnidirectional vision system and to develop various application systems. The system consists of a large number of omnidirectional vision sensors located in the environment and connected with a computer network. Compared to existing systems using standard vision sensors, the distributed omnidirectional vision system can provide a wide scene coverage with fewer vision sensors. In addition, the system can observe an object from various viewpoints, which enables robust recognition of the object. These merits allow us to develop practical vision systems. In this thesis, we study various techniques in the distributed omnidirectional vision system, especially the following respects: methods for measuring the sensor and object locations by observation, methods for measuring the sensor locations by observation without complex numerical expressions, and development of application systems. The first issue deals with one of the most fundamental and important techniques in the distributed omnidirectional vision system. The distributed omnidirectional vision system has different aspects in localization compared to existing multiple camera systems. In addition, various methods should be considered according to situations, e.g., whether the locations of the sensors and/or objects are known. The second issue addresses localization of