SLAM algorithm for mobile robot localization with RGB-D camera

This paper presents a three dimensional reconstruction and map building method for mobile robots. The system is able to evaluate the movement of vehicle in real time and continuously update the necessary maps based on new information with the help of using data of the Kinect sensor. The essence of method is that two three-dimensional reconstruction of environment made in different times are largely overlapped. For evaluation of displacement the algorithm searches cohesive feature point pairs in the three dimensional point clouds of environment. Definition of transformation among three dimensional feature points made in different times is based on a SVD algorithm. Joining can minimize the joint fault and delete the erroneously detected point pairs during a multi-stage iteration. Key–Words: mobile robot navigation, mapping, three dimensional reconstruction, SLAM, Kinect, RGB-D camera, SVD