Experimental Analysis of Measurements Fusion for Pose Estimation Using PMD Sensor

Abstract

This article presents the experimental investigation of the fusion concept of two relative position and orientation estimates of the target using single Photonic Mixer Device (PMD) sensor for the frame-to-frame tracking. For each frame PMD sensor provides co-registered depth and amplitude images of the scene. We propose to use two different pose estimation techniques for each of the data channel with a further fusion of its measured state vectors. The fusion architecture of the state measurements is based on a low complexity weighted average algorithm. The weights for the fusion operator are calculated experimentally with the real data from PMD sensor mounted on DLR's European Proximity Operations Simulator. The fused state vector obtained with experimental results outperforms the accuracy of the two incoming state measurements. This allows us to ensure robust pose estimation of a tumbling target for the whole tracking.