Development of a Robust Ear Recognition Algorithm using Planar Approximation

In recent years, multimodal authentication combining multiple authentication methods such as face recognition, iris recognition, and ear recognition has attracted attention. Ear recognition can achieve nearly 100% recognition accuracy in ideal conditions. However, it suffers from a significant decrease in recognition accuracy in the presence of occlusion, pose variations, and lighting variations. In this study, we focused on the influence of variations in ear poses on recognition accuracy. To address this issue, we employed a method of data augmentation for the training set, incorporating a variety of ear poses using plane approximation techniques. The results confirmed a noticeable improvement in recognition accuracy due to the inclusion of multiple poses.