Ear Recognition Based On Geometrical Feature

Ear recognition is a new technology of biometrics, which has been paid attention to by more and more scientists because of its unique feature and applying angle. Ear alone can serve as a biometrics in some areas, and it can also be combined with other biometrics to act as multimode biometrics. The unique characters of ear make it possible to enrich biometrics technology and to make up the disadvantages of other biometrics. The feasibility of ear acting as biometrics has been proved already.As a specific biometrics, ear has unique physiological structure and position, which are embodied in the difference of helix shape and inner ear structure. According to these features, this paper proposes a novel recognition method based on ear contour geometrical structure, it includes three main phases, ear contour edge detection, helix and concha features extraction and patterns recognition. In the process of ear contour extraction, a new edge detection method based on contour composition is proposed. First ear main contours are derived from canny operator with large scale parameter, and then these contours are reinforced by the detailed edge image which is extracted by canny operator with small scale. This method overcomes the disadvantages of classical edge detection operators, and clear edge image with less noise is gained. In the process of feature extraction, helix shape feature and inner ear structure feature are extracted separately. Helix shape feature vector is composed of several key distances on helix edge, which describes the shape of the whole ear. Inner ear structure feature vector is composed of corners on concha edge, which describes the textural feature of the inner ear. In the process of patterns recognition, matching rules are designed for each vector to identify individuals. And experiment results show that FAR (false acceptance rate) and FRR (false rejection rate) achieve 0.8% and 6.7% respectively, and recognition accuracy can achieves 92.5% from selected image database.At the end of the paper, force field transformation theory is introduced and some experiments on ear images using this theory show the unique features such as potential wells, potential channels and field lines. Such approach allows the successful processing of image RST queries, and the result of experiments approves the certainty and uniqueness of the features. All above proves that the force field transformation theory can be utilized in ear recognition.