Mobil robot localization using RGB-D camera

Abstract

The paper shows a movement tracking and a map building system based on a RGB-D camera for mobile robots. With the data of the Kinect sensor the system estimate the displacement of the sensor. It uses only the color and pixel level depth informations of the camera. Beside the estimation of displacement the three dimensional map of the environment is obtainable known the movement of the sensor. This information can be used by the navigation algorithm of the robot during its mission. The paper describe the operation of the process which based on the measuring of sensor displacement. Value of displacement is based on the join of the three dimensional point clouds of the environment. The point clouds made in different time and place can be overlapped in a manner. Searching of the overlapped feature points happens with the SURF feature detector. The transformation among the smaller set of points (how much moved the sensor in the space during the measurement) is given by the ICP algorithm. The joining happens in several steps, during this the erroneously detected point pairs are eliminated. In this way more accurate results are available.