2P1-F17 SIFT特徴量と物体の相互拘束関係を用いた室内シーン画像の領域抽出とAdaBoostを用いた物体認識

概要

論文の詳細を見る
For the purpose of simultaneous localization and mapping (SLAM) of an indoor mobile robot, a method to process only one monocular image (640×480) of the whole view of environment is proposed. This method imitates that a human can grasp momentarily the situation of a room, i.e., the layout of objects/obstacles. It is not required for a robot to move around the room while probing the objects/obstacles by some sensors, which is usually done in SLAM research. The proposed method is concretely organized as follows: 1) narrowing the searched area using invisible floor marks, 2) extracting features based on scale invariant feature transform (SIFT), 3) template matching using SIFT features with assistance of partial templates and spatial relation to the floor, and 4) verification of object category by AdaBoost classifier using Haar-like features based on semantic information on the object shape. Then, a robot builds a highly structured 3D map.
一般社団法人日本機械学会の論文
2009-05-25