2P1-F05 ステレオ視を用いたオンライン道路境界モデリング

概要

論文の詳細を見る
The estimation of traversable region is one of the most important functions for autonomous mobile robots in outdoor environments. This paper describes a stereo-based estimation method. Matsushita et al. realized a robust estimation of traversable region using two sensors, a laser range finder (LRF) and cameras, and models of unbranched and branching roads in a particle filter framework. Since the LRF can only obtain data on the scanning plane, however, the robot sometimes to failed to correctly discriminate between the two road models. Therefore, it was often observed the situation which can't determine straight road model or cross road model. We thus propose to use a stereo camera which has a wide field of view in place of the LRF. The proposed method evaluates the likelihood of traversable region boundaries using brightness, color, and height information obtained with a stereo camera and can robustly and efficiently estimates the traversable region by switching the road models by referring to the trends of the likelihood values.
一般社団法人日本機械学会の論文