抄録
The wide-angle fovea (WAF) sensor comprises a specially made wide-angle fovea lens and a commercially available CCD/CMOS camera with photosensitive elements of uniform size. The sensor realizes a 120-degree-wide field of view (FOV) and high magnification in the central FOV without increasing the number of pixels per image. This paper focuses on the multifunctional use of an input image with space-variant spatial resolution that enables an autonomous mobile robot to avoid obstacles during locomotion. In order to use the WAF-input image efficiently, image processing for central vision, i.e., detection of 3D obstacles, and image processing for peripheral vision, i.e., self-localization of the mobile robot, are performed simultaneously and cooperatively. The comparison of the simulation results of spatial resolutions of the WAF lens and a pinhole camera (PHC) lens shows that the WAF lens can be used for depth measurement in the central FOV and self-localization in the peripheral FOV by the parallel stereo method and the two-parallel-line algorithm, respectively. The results obtained by the WAF lens are more accurate than those obtained by the PHC lens. Autonomous locomotion of the mobile robot has been demonstrated by performing two obstacle avoidance experiments.
