3D Ear recognition using rigid point set registration and coherent point drift algorithm

In this paper, we present an approach to recognition3D ear based on the coherent point drift (CPD) method with rigid point set registration and EM optimization algorithm. The proposed work automatically detects the pit point and nose index from the 3D ear image. The corresponding features of the detected points of the input 3D ear are passed to the CPD algorithm for alignment and matching of detected points. The coherent point drift (CPD) algorithm works fine with rigid and non-rigid point set registration. The proposed algorithm used the CPD method for rigid point set registration. Since the anatomical structure of the ear consists of many intrinsic unique feature points, the alignment between two such point sets becomes easier to recover the transformation from one point set to the other. The performance of the proposed work has been validated on the UND collection J2 ear dataset which is unconstrained in nature. Here, illumination changes, pose variations, occlusion by earrings, rotation, and hair are included in this dataset. So these significant challenges are overcome by the proposed algorithm. The experimental result on these datasets shows that the proposed method is effective and feasible under varying environmental conditions. The result of the proposed algorithm shows that CPD based approach can accurately identify a person based on their 3D ear point data and compete with the other state-of-the-art method.