Researching On Finger-vein Image Classification And Recognition Based On Hierarchical Framework

As a new biometric technology,finger-vein recognition has attracted more attentions by researchers since this trait has liveness, innerness, and non-invasion etc. Currently, the efficiency and accuracy of finger-vein image identification and retrieval over a massive database are still two open problems. Finger-vein classification is beneficial to both reducing the time cost and improving the efficiency of recognition. So, the main purpose of this thesis is to explore some basic issues for finger-vein image classification based on the finger-vein imaging properties.A new finger-vein classification strategy based on a hierarchical framework is proposed.Since there is no obvious prior categorization information for finger-vein image, an unsupervised clustering scheme is applied. First, according to the characteristic of finger-vein image appearance feature, the imaging qualities are utilized for the first layer finger-vein feature representation. Second,considering the differences of vascular network in complexity,the image content features are adopted for the second layer finger-vein feature representation.Here, the content features include the invariant moment features and the eight channels Gabor filter statistics information features. Thirdly, an improved k-means clustering algorithm and SVM classifier are applied for automatically classifying finger-vein images,the experimental results demonstrate that the proposed method has good performance in finger-vein classification.Finally, POC(Phase-Only-Correlation) algorithm is used to evaluate the matching performance of the finger-vein based on our method. Experimental results show that the proposed method has a good performance in improving the recognition efficiency as well as recognition accuracy.