Low-cost sensor-based exploration in home environments with salient visual features

摘要

This paper describes an exploration method based on sonar sensors and a stereo camera. To build an accurate map in unknown environments during exploration, SLAM (Simultaneous Localization and Mapping) problem should be solved. Therefore, a salient visual feature (SVF) extraction method is proposed for SLAM. The key concept of SVF extraction method is to extract meaningful features of environments using SIFT keypoints. The extracted SVFs are applied to the EKF (Extended Kalman Filter)-based SLAM framework. This proposed method was verified by various experiments which show that the robot could build an accurate map autonomously with sonar sensors and a stereo camera in various home environments.