2P1-F14 移動プラットフォーム搭載カメラによる移動障害物検知

概要

論文の詳細を見る
In this research, we aim at moving obstacle detection. We operate frame differencing between consecutive image frames which most commonly-used method to detect moving object. But straightforward frame differencing operation is unsuitable when camera is moving. The method proposed by B. Jung and G. Sukhatme solve this problem by converting image to adjust the background of a consecutive images. The accuracy of this method's conversion relies on accuracy of correspondence used to calculate conversion matrix. In this paper, we propose a technique which makes conversion accurate. In order to achieve the objective, divided feature detection, Least Median Squares, judging distance between correspondences are introduced. As a result, it is shown that the accuracy of conversion is improved by using these methods. Furthermore, we mark moving obstacle using the result of frame differencing operation.
一般社団法人日本機械学会の論文
2009-05-25