Study On Omnidirectional Vision System For Mobile Robot Self-Localization And Navigation

The research, described in the thesis, concerns the self-localization and navigation of mobile robots by target recognition and tracking under the artificial targets environment. The thesis emphasizes the three aspects of calibration and correction for omnidirectional vision, target recognition and tracking, and localization and navigation of mobile robots.Calibration and correction for omnidirectional vision is an enabling key for localization of mobile robots. We use a fisheye lens upwards with the view angle of 185°to build the omnidirectional vision system. Even the fisheye lens takes the advantage of an extremely wide angle of view; it will bring an inherent distortion in the image and this distorted image must be rectified or be restored. Four methods for calibration of a fisheye image are proposed. The calibration parameters are employed for the correction of image distortions. Three imaging rules are conceived for the designs of fisheye lenses. The regulations are discussed respectively and the distortion correction models are generated. An integral distortion correction approach based on these models is developed. And the correctional effects are shown in real-time process.Target recognition and tracking is the foundation for localization of mobile robots. The thesis introduces three target recognition solutions: they are color and area recognition, character parameter recognition and line recognition. In the integral experiment, the character parameter is used, because of its high precision and the capability of eliminating yawp. Line recognition is used in line equation localization alone. The thesis also introduces two tracking methods: color extension and particle filter tracking. Because of the huge calculation and worse real-time process of particle filter, color extension is used in our experiment. In the experiment, we find that character parameter recognition with color extension tracking is the correct choice, and can satisfy the precision and real-time process in the indoor environment.Three localization methods are proposed, including two points localization, world localization and line equation localization, and are all accurate by experiments. Two points localization is chosen in our integral experiment, because of its simple algorithm and high precision. The thesis also introduces two navigation methods, including basic navigation and PID control navigation. In our integral experiment, we use basic navigation to get angle and speed, then drive robot moving by PID control.