A panoramic stereo-vision detection and recognition system

Abstract

In this paper, a panoramic stereo-vision detection and recognition system based on stereo-vision detection and recognition techniques is proposed. The study builds a panoramic image capturing module including four cameras and a fixed camera foundation for capturing panoramic images and develops an object detection technique for panoramic images. In addition, it also proposes a stereo-vision detection algorithm for non-coplanar images. In this study, in order to get stereo-vision information, the panoramic image capturing module uses wide-angle lens to obtain more overlapped areas of images. Overlapped areas are used to calculate the disparity of captured image in stereo-vision system. However, images become distortion when using wide-angle lens, the image distortion increases the difficulties of the disparity calculation and the panoramic image stitching. Therefore, the distortion correction procedure is necessary.