Ear Detection and Recognition in 2D and 3D

Abstract

Although in the field of computer vision and pattern recognition ear biometrics has received scant attention compared to the popular biometrics such as face and fingerprint, ear biometrics has played a significant role in forensic science and its use by law enforcement agencies for many years. Iannarelli [8] developed the famous ear classification system based on manual anthropometric measurements of ear photographs of thousands of people. The “Iannarelli system” is pdfed upon 12 measurements illustrated in Figure 2.1. The locations shown are measured from specially aligned and normalized photographs of the right ear. This system has been in use for more than 40 years, although the reliability of this earprint-based system has recently been challenged in court [20, 21]. All the work described in the previous paragraphs was done manually. Rutty et al. [22] considered how Iannarelli’s techniques might be automated. Meijermana et al. [23], in a European initiative, have looked into the value of earprints for forensics. The early research on automated ear recognition was focused on 2D intensity images. However, the performance of these systems is greatly affected by imaging conditions. It is possible to image an ear in 3D. The 3D data collected by a range sensor provides geometric shape information that is relatively insensitive to illuminations. Design of a biometrics system based on 3D data has been proposed in recent years. Furthermore, the ear can be thought of as a rigid free-form object. Therefore, in this chapter the related work is reviewed in several areas: 2D ear biometrics, 3D biometrics, and 3D object recognition.