A FLEXIBLE GRAPH MATCHING-BASED APPROACH TO RECOGNIZE CATTLE MUZZLE PATTERNS WITH GROWING DEFORMATIONS(International Workshop on Advanced Image Technology 2006)

概要

論文の詳細を見る
In this paper, we present an improved method to recognize biological textures, especially cattle muzzle patterns. Biological textures are textured patterns on living organisms such as human fingerprints or cattle muzzle patterns. It is widely known that these biological textures can be used for identification, as an example, human fingerprints can be matched nearly 100% by extracting feature points called minutia. However, unlike human fingerprints, cattle muzzle patterns are structurally complicated and have also pattern deformations during the growing stage, and thus they cannot be recognized by just applying a conventional recognition method of human fingerprints to them. From this viewpoint, the authors proposed a recognition method for biological textures in the previous paper[1], which is able to flexibly deal with growing deformations. Although the method performed well to some cattle muzzle patterns, there was still plenty of room for recognition performance. This paper describes improved points to make our previous method more useful and robust. Our main improved point is to additively incorporate muzzle lines-based graph matching method in the recognition stage. We show the results of evaluation experiments to verify the effectiveness of our improved method.
一般社団法人電子情報通信学会の論文
2006-01-02