A Research Of Fingerprint Image Recognition Algorithms

With more and more electronic devices coming into our life, there is an impending need of more convenient and safer personal identification techniques. The aim of our research is to develop an automatic fingerprint identification system (APIS) for personal identification.In this paper, we present an intensive research in the algorithm of fingerprint recognition. In China, it seems no detailed paper published in fingerprint recognition, so it is a difficult and hard work for us. In our research the following topics are involved:Computing the orientation image of a fingerprint image. Each pixel in the orientation image stands for the local ridge orientation of a fingerprint. It is the intrinsic characteristics of the fingerprint image. After an intensive analysis of several methods of computing the orientation image, we select the "Least Square Estimate" method in our research.Image enhancement. We improve the clarity of the ridge and furrow structures of the fingerprint image to facilitate automatic feature extraction. First, we introduce the fingerprint image enhancement algorithm based on Gabor filter. Then we propose our fingerprint image enhancement algorithm based on local ridge orientation.Feature extraction. According to the FBI's recommendation, we extract the ridge's ending and ridge's bifurcation as the representation of the original image. We record the feature's x, y coordinates and it's orientation for the purpose of fingerprint matching.Minutiae match. Improving Anil Jain's algorithm, we propose a minutiae match algorithm. Our improvement is focused on the following two aspects: First, we use a minutiae alignment algorithm based on the iteration of the neighbor minutia centered at the minutia gravity. Secondly, we use a modifiable bounding-box to carry out minutiae matching, which makes our algorithm more robust to the non-linear transform of the fingerprint images.Fingerprint classification. According to the number and relative position of "core points" and "delta points" extracted from the fingerprint image, we classify the fingerprint into the following six classes: arch, tented arch, left loop, right loop, whorl, twin loop. Aiming at the condition of losing some delta points, we bring forward a new method of secondary fingerprint classification. By the new method, we can classify a nice bit of fingerprint images that cannot be classified by the former method.A lot of experiments have been done to test our fingerprint identification and classification algorithms. The experiment results show that our algorithms perform very well with high identification rate and speed. They show also that our algorithms are robust for some poor quality fingerprint images.