High Resolution Fingerprint Matching Method Based On Deterministic Annealing Algorithm

Nowadays, the automatic fingerprint identification systems(AFIS) are widely used in many fields, such as fingerprint attendance, fingerprint password, even in the some social test. So the weaknesses of AFIS are revealed gradually. Some criminals using the cheap fingerprint set to steal the user’s fingerprints, the information and property security of user’s are treated. In the face of the existing defects of the AFIS, high resolution fingerprint identification technology(HRFIT) becomes a hotspot research. HRFIT has the characteristics of fraud prevention, high accuracy, more feature points etc. But there exists some defects. First, high resolution finger prints are much sensitive to nonlinear deformation. Second, the number of sweat pores that extracted is large, so it’s time-consuming for sweat pores matching. Third, the local characteristics of fingerprint are amplified. Last, because of the influence of the translation and rotation, may be two matching parts of fingerprint are not the same area, this has much effect on the result of fingerprint identification. In this paper, we mainly study the matching problem of high-resolution fingerprint characteristics.In the process of fingerprint gathering, it is inevitable that the translation and rotation will be occurring, so the calibration is the key step before high-resolution fingerprint matching. The calibration method we proposed in this paper is based on the singular points of the fingerprint. We use the type of singular points and the direction information of each singular point to select out the datum points for calibration. Through most of experiments, we get that this algorithm can calibrate fingerprint accurately and effectively, it improves the accuracy of the high-resolution fingerprint identification greatly.One of the hard problems of high-resolution fingerprint identification is the matching of feature points. There has hundreds of sweat pores in each fingerprint image, to filter out the stable sweat pores becomes the key step. In this paper we proposed two methods for high-resolution fingerprint matching: 1) the research of high resolution fingerprint identification algorithm that based on the fingerprint blocking and deterministic annealing technology. By aligning the fingerprints, selecting blocky and finally using the Random Sample Consensus(RANSAC)algorithm for matching. This method effectively reduces the influence of local deformation to the results; 2) the research of high resolution fingerprint identification algorithm based on convex hull of minutiae points and deterministic annealing technique. By the process of aligning the fingerprints, matching the minutiae points and constructing the convex hull, registering by deterministic annealing etc. We improved the problem of migration and rotation effectively, and solved the phenomenon that no singular points in one fingerprint, which effects the results seriously in algorithm one.The results of experiments show that the fingerprint alignment methods we proposed, which can calibrate the fingerprints effectively and can. The two fingerprint matching methods we proposed that have high robustness, and solved the issue that high-resolution fingerprint identification is sensitive to local deformation. The sweat pores selected out by the improved deterministic annealing method for matching are only occupied 20%~40% of the original sweat pores, and can improve the accuracy of high resolution fingerprint recognition rate greatly.