3D Ear based human recognition

Abstract

Biometrics involves automated technology that recognizes an individual using physiological and behavioral characteristics. Classical authentication methods focused on passwords, identity cards, and so on, have certain drawbacks, such as forgetful passwords and stolen or fake identity cards. Biometrics provides a safe way to recognize individuals and circumvent traditional methods. An individual possesses information and it is difficult to steal or counterfeit in terms of biometric recognition. In the past, many biometric features have been investigated. In recent times, ear biometry has attracted significant interest among researchers in comparison to other well-known biometric traits, such as the face, fingerprint, iris, etc. Significant ear recognition works have been conducted using 2D ear as a biometry to authenticate the individual. Nevertheless, in the presence of posture, lighting and scaling changes, the efficiency of 2D ear recognition deteriorates. These issues are addressed by 3D ear images where the rich geometric information inherent in 3D ear images enhances the efficiency of ear recognition. Registration based algorithms are mostly used in 3D ear matching, while features based 3D ear recognition are another approach to matching. Since the ear is highly similar in nature and the interclass differences are very small, it is difficult to design ear recognition descriptors. In 3D ear biometrics, to match two ears, feature based recognition techniques are combined with registration algorithms to match the ear pairs. Primarily, the work presented in the thesis addresses recognition issues related to 3D ear biometrics. It proposes efficient descriptors for 3D ear recognition. The recognition performance of these descriptors are superior or on par with the state-ofthe- art techniques exists in the literature. The experimental analysis are conducted on University of Notre Dame (UND)-J2 collection database to show the effectiveness of the proposed descriptors. Further, an analysis has been carried out to discover the invariant nature of ear and compared it with that of face using 2D and 3D images. The complete age invariant analysis has been conducted using IIT Indore database.