Research On Finger Vein Recognition Technology

Biometrics has a high research value and the broad application prospects in the field of the authentication. Finger vein recognition, identifying people by their finger vein, is a very important research field in biometric recognition. It has attracted more and more attention because of its unique advantages such as living body identification, internal features, non-contact, and easy use. And it gradually entered all areas of human social life, to meet the challenges of the new era.However, in practical applications, subject to the conditions of equipment, the captured finger vein images usually show low quality, it is also difficult to extract the features from those low quality finger vein images. For these above mentioned, it would affect the accuracy of the recognition system. To solve these problems, we do some researches on the feature extraction and the recognition of finger vein image. In this paper, we propose two novel algorithms, one is about the feature extraction of finger vein patterns based on the improved direction detection; the other is about the finger vein recognition based on the two directions of modular two-dimensional principal component analysis. The main attributions in this paper are as follows:1. The principle of finger vein image acquisition is introduced firstly, then the relevant processing technology of finger vein image is analyzed, including the extraction of its ROI, scale and gray scale normalization, image segmentation and image post-processing. Finally, we introduce two kinds way of matching recognition, and the performance parameters of the system.2. A novel feature extraction of finger vein patterns based on the improved direction detection is proposed. Firstly, a modulated Gaussian low-pass filter is used to smooth and enhance the images. Secondly, considering the valley-shaped feature of the finger vein, a new group of oriented operators are designed to detect the directions of a finger vein image and calculate every pixel's directional value accurately. Finally, enhanced image is segmented and refined, and we get the feature extraction. In order to prove the effectiveness of the proposed algorithm, the modified Hausdorff distance is adopted to realize the recognition of the finger vein refined image matching. The experimental results show that the algorithm can not only effectively extract the finger vein texture features, but also achieve the higher identification precision and practicability in matching recognition.3. The classic 2DPCA and 2DLDA algorithm are introduced in this part firstly. Then do the experiments about some combined two-dimensional linear analysis recognition algorithms, such as 2DPCA+2DPCA,2DPCA+2DLDA,2DLDA+2DLDA and do the analysis about the relationship between these algorithms and the number of training samples with the dimension of mapping features. Meanwhile, the main content and steps of modular two-dimensional principal component analysis (M2DPCA) is given in detail, and we propose a new finger vein recognition algorithm based on the two-directions of modular two-dimensional principal component analysis. It makes the finger vein image modular firstly. Then it processes the image using 2DPCA to utilize the local details effectively. Finally,2DPCA or 2DLDA is used to reduce the dimension of features of finger vein image. The advantage of this step can decrease the computational cost and improve the recognition rate of our recognition algorithm. Experiments with M2DPCA+2DPCA and M2DPCA+2DLDA show that our algorithm can achieve a good performance.