Study Of 2-D Skull Image Recognition Based On Contour Information

The biological recognition technology has been a hot research region in recent several decades. This identification technology depending on the features of fingerprint,iris,face has been widely and successfully used on access system,security system and other occasion needing identity witnesses. Because of the limited efficiency of a single biological feature,the multimode features integration is the hotspot that many researchers focus on. Skull recognition that is a new technology has attracted more and more attention in recent years. The main target of this paper is the skull. We extract features of skull's contour to identify individuals. During process of feature extraction, we do a lot of research about some algorithms and a number of experiments. The main work is as follows:Firstly, on the basis of researching lots of references and relevant research in and abroad, we summarize the purpose and signification of the skull recognition. After the pretreatment of the skull image, we choose one suitable way to extract the contour through many different ways. Then, we extract features with different traditional methods to describe the contour curve of the skull. After that, we analyze and compare the results of many experiments.Secondly, a boundary description function of polar-radius mean difference is researched. We can acquire a new chebyshev descriptor through combining the function of polar-radius mean difference with chebyshev orthogonal polynomial. The experiments show that the chebyshev descriptor has the invariance properties of translation, rotation and scaling. The test shows the ability of chebyshev descriptor against noise. We can use the chebyshev descriptor to extract features of the skull contour and have a good recognition result.Thirdly, according to the theory of moment invariants, we propose a new approach of the homocentric polar-radius moment to describe the boundary curve of the skull. We divide the whole boundary into some small regions with several circles which have the same center and different radiuses. The center of all the circles is the centroid of the object boundary. In this way, we can acquire a group of vectors as the homocentric polar radius moment to describe the skull boundary. All the theories and experiments show that the homocentric polar radius moment has the invariance properties of translation, rotation and scaling. It has a good ability to recognize the similar images against noise and to describe the skull boundary. Finally, several classification methods are discussed. We select a best method among these ways through lots of experiments for classification and recognition.We do in-depth research to find methods for extracting features from the boundary of the skull. After comparing with many traditional algorithms, some new methods are proposed. Through a mass of experiments, the stability of the methods is proved. Also, we can conclude that it is a useful way for biological recognition technology by skull recognition.