Research Onalgorithm Of Fingerveinrecognition

The finger vein recognition technology is a new biometric identification technology, which originated from the finger vein near-infrared imaging technology. Finger vein recognition technology has the advantages of living body recognition, internal features, uniqueness, high level of security and non-contact. Therefore, it has become the important biological recognition technology in our life.According to the recent research of finger vein identification technology algorithm, this paper proposes two novel algorithms for finger vein recognition. One is personal identification based on finger vein minutiae features, and the other is personal identification based on wavelet decomposition and KECA. Main contribution and innovative points are as following:1. Finger vein image preprocessing, which includes the extraction of ROI, size of the finger vein normalized, gray scale normalization and image enhancement. After the finger vein image preprocessing, it can reduce the redundant information, highlight the target area.2. Finger vein has the bifurcation points and ending points, but they are unable to completely characterize the information of the image andhave influence on the recognition rate. In order to overcome the inadequate minutiae points from finger vein for matching, especially in non-ideal images, this paper proposes to integrate the minutiae features that are extracted by SURF algorithm. Then, the two kinds of feature points obtained by these two algorithms are integrated and the Euclidean distance is used to integrate these two sets of feature points, and GA is used to select the best sets of feature points. At last, the modified Hausdorff distance(MHD) is utilized to analyze the spatial similarity between the minutiae features sets for matching.3. If the kernel entropy component(KECA) is used directly on the finger vein image, not only its ability to extract the feature classification is poor, but also the calculation amount is very high. So this paper presents wavelet decomposition algorithm. The low frequency image after wavelet transformation has high energy. It also reduces the noise, and the image size is only 1/4 of the original image. KECA is used in the low frequency image which can greatly reduce the amount of calculation and the Euclidean distance classifier can be used for match.All experiments were performed on MATLAB platform. The database is provided by Tianjin Key Laboratory of Finger Vein. The database includes 64 fingers samples, each sample containing 10 finger vein images. The total of the finger vein image is 640 and the size of image is 320 × 240. The experimental results show that the recognition rate of personal identification based on finger vein minutiae features reaches to 99.56%. The recognition rate of personal identification based on wavelet decomposition and KECA reaches to 98.9%.