IMPLEMENTING AND DESIGNING BIOMETRIC SYSTEM FOR DETECTION AND IDENTIFICATION WITHOUT LIMITATION OF HUMAN EAR

Abstract

In the current article, by designing and implementing an ear-based biometric system, ear recognition operations in images and identification with the human ear are performed automatically and without the need for user interaction. To design the system, in the ear detection step, in the main images of the database, the Image Ray Transform (IRT) algorithm is used, and also a method is proposed for normalization. In the identification stage, local tissue descriptors of Local Binary Pattern (LBP) and Binarised Statistical Image Features (BSIF), and Weber local descriptor (WLD) were used to extract the feature and the nearest neighbor k algorithm was used for decision making. The result of the accuracy of applying the binary pattern feature on the ear images is equal to 50%. The result of the accuracy of applying statistical properties on binary images is equal to 93.88% and the result of the accuracy of applying Weber local descriptor feature on ears images is equal to 60.05%. The method proposed in this paper improves the performance of the Weber local descriptor (WLD).