Vision Based Mapping and Localization in Unknown Environment for Intelligent Mobile Robot

Simultaneous Localization and Mapping (SLAM) is a key component of mobile robot's navigation. In this paper, we present a vision-based system of mapping and localization. The system builds a map which contains 3D landmarks of environment. 3D landmarks are reconstructed based on sequential Harris Corner Features. In order to match regained features with landmarks in the map, we present a simple method, i.e. Inverse Projection. The method is relatively simple and effective. We also propose an efficient observe model that simplifies the Jacobian Matrix when apply the Extended Kalman Filter (EKF) framework. We conduct experiments in both simulation and real-time environment, and give error analysis for robot's and landmarks' position. Experiment results show that landmarks are localized accurately and robot trajectory is well estimated by matched landmarks.