Seal registration and identification based on SIFT

At present, most of the existing seal image registration methods are based on shape, and most of the methods can´t tolerate the image scale change. To overcome these problems, a general seal registration and identification approach is proposed in this paper. This method uses the classic scale invariant feature transform (SIFT) algorithm to extract the key points. To register the seal accurately, the Random Sample Consensus (RANSAC) algorithm and the Least Squares Method (LSM) are used to obtain the transformational matrix. Then the invariant features are extracted based on residual image. At last, the Normal Bayesian classifier (NBC), K-Nearest Neighbor classifier (K-NN) and the Support Vector Machine (SVM) are combined to classify the feature vector. The experiments demonstrate the identification ability of the proposed approach.