Combining PMD and Stereo camera for Motion Estimation of a Mobile Robot

This paper introduces an improvement to the 6D motion estimation solution by combining the Photonic Mixer Device (PMD) and Stereo Camera. The main feature of the stereo camera is the higher resolution of the 2D image compared to the one from the PMD camera. Whereas the depth information derived from a PMD is usually far superior to the result from a stereo camera. This work hence proposes a combination scheme for the PMD and stereo camera in order to improve the results of the motion estimation. The combined setup was placed on a mobile robot and carried on the 6D motion estimation task using a provided artificial landmark. The robot was moved around while the combined setup simultaneously captures 3D images of the landmark from each new position. The motion was estimated based on the matching of the captured 3D images between two successive positions. The classical singular value decomposition (SVD) algorithm was used to solve the matching problem. The referent points for the SVD algorithm were extracted from the landmark using a robust corner detection algorithm. The experiments were fulfilled using 1) stereo 2) PMD and 3) combination of stereo and PMD camera. The results of these three arrangements are compared and the outcome of the comparing are presented.