Abstract
In 3D object recognition study, most of the existing method relied on model-based method. However, it requires detail of angle information in learning process or complexed model description. In this paper, we present a novel method which does not model-based but appearance-based approach in order to avoid those problems. This proposed method recognizes and presumes the object area by voting process and appearance estimation. Then, it removes false recognitions by comparing the estimated appearance with the recognition area. The experiments showed that the performance of our method outperforming other works.
