A Novel Omnidirectional Stereo Vision System via a Single Camera 1

Obtaining panoramic 3D map information for mobile robots is essential for navigation and action planning. Although there are other ways to fulfill this task, such as ultrasonic sensors or laser range finders, stereo vision system excels them in its precision and real-time speed without energy emission.But the conventional stereo vision systems are limited in their fields of view (FOV). An effective way to enhance FOV is to construct an omnidirectional vision system using mirrors in conjunction with perspective cameras. These systems are normally referred to as catadioptric and have been applied to robot localization and navigation by several researchers (Bunschoten & Krose, 2002; Menegatti et al., 2004). A common constraint upon the omnidirectional sensors modeling requires that all the imaged rays pass through a unique point called single viewpoint (SVP) (Baker & Nayar, 1999). The reason a single viewpoint is so desirable is that it is a requirement for the generation of pure perspective images from the sensed images. These perspective images can subsequently be processed using the vast array of techniques developed in the field of computer vision that assume perspective projection. The mirrors popularly used to construct wide FOV catadioptric are hyperbolic or parabolic. But the latter must be coupled with expensive telecentric optics which restricts them to limited applications in panoramic vision. Mobile robot navigation using binocular omnidirectional stereo vision has been reported in (Menegatti et al., 2004; Yagi, 2002; Zhu, 2001). Such two-camera stereo systems can be classified as horizontal stereo systems and vertical stereo systems according to their cameras’ configuration. In (Ma, 2003), the cameras are configured horizontally and the baseline of triangulation is in the horizontal plane. This configuration brings two problems: one is that the epiploar line becomes curved line leading to increasing computational cost; the other is that the accuracy of the 3D measurement depends on the direction of a landmark. In the omnidirectional stereo vision system (Gluckman et al., 1998; Koyasu et al., 2002; Koyasu et al., 2003), two omnidirectional cameras are vertically arranged. Such configuration escapes the shortcomings brought by horizontal stereo system, but the power cable and data bus introduce occlusion to the images captured by this configuration. In