Finger knuckle print recognition based on SURF algorithm

This paper proposed a knuckle print recognition algorithm based on SURF algorithm. Firstly, a coordinate system is defined based on the local convex direction map of the finger knuckle print(FKP) to align the images and a region of interest (ROI) is cropped for feature extraction; secondly, the key points are extracted with fast Hessian detector, to which an orientation was assigned according to the Haar wavelet responses inside the neighbor circle area of the keypoint; and an orientation invariant descriptor is constructed for each key point. In recognition, testing FKP features are matched to the template features for initial correspondences, then random sample consensus (RANSAC) is employed to establish the geometric constraint which is used to remove false matching. The amount of final matched point pairs is referred to decide the consistency of two palm images. Plenty of experiments show that FKP can be recognized with high accuracy. The method is invariant to rotation, scale and viewpoint changes, which proves its robusticity.