Robust feature extraction for 3D reconstruction of boundary segmented objects in a robotic Library scenario
In 2010 IEEE/RSJ International Conference on Intelligent Robots and Systems, (IROS), IEEE, pages 4540-4547, 2010.
In this paper a vision system for robust feature extraction and 3D reconstruction of boundary segmented objects is presented. The goal of the system is reliable perception of a professional life environment in a scenario of the rehabilitation robot FRIEND. Reconstructed scenes are used to plan object manipulation with a 7-DoF manipulator arm. The robustness of boundary feature extraction is achieved by the means of including feedback control at image segmentation level. The objective of feedback is to adjust the segmentation parameters in order to cope with scene uncertainties, such as variable illumination conditions. Robustly extracted 2D object features are provided as input to the 3D object reconstruction module of the FRIEND vision system. The performance of the proposed approach is evaluated through experiments in the Library scenario of the robotic system FRIEND.