Image matching by eigen template method for multi-class classification

Abstract

An Image matching technique of target objects recognition and detection is widely used in industrial image processing. In this paper, the authors proposed eigen template method for two dimentional target objects recognition and detection. We have proposed eigen template method of applying the principal component analysis (PCA) to image matching. Also, the authors have proposed to edge based eigen templates method for robust and efficient image matching. These methods can estimate target objects position and pose from two dimensional images. In this paper, eigen template method is extended to the method that can classify multi-class target objects. Our experiment showed that the proposed method can recognize multi-class target objects and estimate position and pose. And the authors showed that the proposed method can efficient image match compared with the previous methods.