On Object Perception in Enhanced Reality Environment for Robot Telecontrol
Chensheng Wang, Fei Wang, Tjamme Wiegres · 2009
Object recognition is one of the hot topics in computer vision. Existing techniques for object recognition are based on image processing, which tends to be incapable to recover the topological information of the object. In this paper, based on the analysis of human perceptual habit, a novel strategy for object recognition is proposed. The method perceives an object in the scene by means of existing shape knowledge coupling. And a mechanism is designed to grow the system intelligence by depositing the shape knowledge into a repository once a new object is encountered. This makes the system be in a constant way to become more and more intelligent. In addition, the proposed strategy is advantageous in rebuilding the whole object information with the support of the repository. Experiment results carried out in the enhanced reality environment for robot telecontrol shows that the proposed strategy is both valid and effective for the designed application.