1P1-L15 ステレオカメラを持つ移動ロボットによる屋外自己位置推定(移動ロボットのための視覚)

概要

論文の詳細を見る
This paper describes a method of outdoor localization which only uses a stereo camera. To obtain a wide field of view, we mount a stereo camera on a pan-tilt head. 3D information from stereo is projected onto a robot-centered 2D grid map. Then building areas are extracted with a support vector machine (SVM)"-"based classifier. The map given to the robot is a 2D map which describes building shapes by lines, and can be edited. We use a particle filter for localization. For each particle, we first apply a hidden line removal to the input map by visibility checking, and then calculate the likelihood of the particle by matching the map with the local map. The proposed method is successfully applied to real outdoor data.
一般社団法人日本機械学会の論文
2011-05-26