Object classification using dempster-shafer theory

Abstract

This paper presents a new method of complex object recognition and classification by a mobile robot. Through applying Dempster-Shafer theory, our algorithm allows to make use not only of object features, but also of information about sensor uncertainty, possible occlusions and can handle contradictory evidence in an integral way. The proposed method has been implemented and tested with complex objects from a real indoor environment, showing that it can be efficiently applied for cluttered data and also make correct decisions even when whole parts of the objects are occluded and the visible parts lack unique features. © Springer International Publishing Switzerland 2014.