Moving surface extraction based on unordered hexagonal perfect submaps projection: Applications to 3D feature tracking

In this paper, we present a structured light technique based on the projection of a color coded hexagonal array that is able to obtain range images from moving scenes. Repetition and disorder is allowed in the codeword which implies several advantages: the mean Hamming distance between contiguous codewords of the pattern increases, the code loss due to occlusions and discontinuities can be efficiently handled and the computational cost in the pattern-image correspondence phase is highly reduced.The structured light projection system has been tested under real moving scenes on medium resolution range images and for slow controlled movements. In order to validate the performance of our range vision system we have used it to identify and track several 3D feature points as the scene moves. To measure the accuracy of the tracking a 6 DOF manipulator robot has been included in the experimental setup. All this experimental work, the results and the main contributions of our method compared to other perfect map and submap based techniques are detailed in the paper.