Evidence Selection for Object Recognition and Pose Estimation based on Bayesian Framework

Abstract — In this paper, a methodology to select proper evidence set in a framework for 3D object recognition and its pose estimation is proposed. To recognize and estimate 3D object pose accurately, photometric and geometric evidences such as color blob, SIFT points and lines, can be utilized as single or multiple features in a sequence of images. However, to guarantee dependability in visual perception the system have to cope with environmental variation that includes change of illumination, amount of texture, and distance to object. So, we made monitoring system to observe the change of environment. And main contribution of this paper is to develop and improve the recognition strategy by proper evidence selection by using Bayesian rule that can be working robustly in various environmental conditions. The experimental results with a single stereo camera show the feasibility and effectiveness of the proposed method in an environment containing both textured and texture-less objects. I