Research On Algorithm For Feature Detection And Matching Of Arm Vein's Infrared Image

Unlike traditional vein recognition technology which are limited to hand area(finger, hand dorsal, palm), a new biometrics recognition technology: arm vein recognition technology is proposed in this paper. Compared with hand, arm has larger area, more veins and abundant features, which makes it has a broader application prospect. The studies of vein recognition in these years are analyzed, and then new feature detection and matching algorithm of arm vein is proposed in this paper. The main works are as follows:First, single-pixel wide vein lines and outlines are extracted from the near-infrared arm images, and then a smooth vein skeleton structure is obtained after trimming burrs in images. The orientation and scale normalization algorithms are studied based on the coordinates of all the points of the outlines. Finally, the characteristics of the vein lines of arm are analyzed and the vein lines are segmented into curve segments. In the segmentation, bifurcation parts of vein lines are detected to segment three intersecting lines and inflection points are detected based on the chain code representation of curves to segment lines with radical inflection points.Then, to study every curve segment, the closest direction is calculated as the direction feature based on the chain code representation of curves. The influence on direction of selecting different initial point and the accuracy of direction are verified by experiments. The position feature is studied based on the results of orientation and scale normalization, and then the separability of the position feature is verified. The Hu invariant moments are extracted as the shape feature. The reliability and separability of Hu invariant moment when faced with rotation and scaling are analyzed and verified.At last, a matching algorithm for curve pairs is studied, with which the same structure of the two matching images can be found. The coordinates of all the points in matching curve pairs are the feature space, modified Hausdorff distance is the similarity measure, and the optimal space transformation of matching curve pairs is calculated by using PSO method. And the matching ratio is computed based on the overlap ratio of all the transformed vein points. The experiments show that the proposed algorithm makes a better recognition rate, and it proves the arm vein as a new biometric feature, is an effective feature for identity authentication.