Knowledge-based Incremental Bayesian Learning for Object Recognition

Abstract

Some of object recognition approaches are very<br/>effective in environments such as industrial settings, where the<br/>position and orientation of object could be controlled. However, in everyday human environments, objects are not located in the same place at all times; rather, they are cluttered in such a way that some of them are visually occluded. Thus, <br/>this paper proposes a method of robust object recognition combing<br/>ontology and probabilistic inference. The basic idea even in<br/>a human environment there is organizational principles that<br/>objects are co-occurred with their related objects. This enables<br/>a robot to recognize object dependably. To demonstrate the<br/>benefits of the proposed approach, a case study is conducted<br/>in a human working environment.