Identification System Based On Hand-dorsa Vein Pattern

As the increasing demand of social security and improvement of identification technology, biometric recognition is becoming more and more important and usable in computer vision as well as in pattern recognition. Hand-dorsa vein pattern with many excellent characteristics, such as uniqueness, stability and hard to copy or forgery, is developed considerably. Although this technique had been put forward to product, there is a lot of work to do:optimizing device design, customizing image segmentation algorithm, optimizing feature extracting and recognition method.The design and realization of vein pattern identification system would be discussed in this dissertation, based on the study on parts of the system.Firstly, design and realization of vein image capturing device is optimized. In order to maximize the quality of captured images under the relatively low cost, the procedures of prototype, design, trial-produce, improvement and finalization are gone through with infrared imaging theory. A low cost capturing device making use of infrared LED arrays has been made up successfully.Secondly, the customized image pre-processing methods are discussed. After ROI is extracted using centroid as centre, simple real-time pre-processing methods, like filtering and enhancement, are tested.Thirdly, a gradient-based image segmentation algorithm is compared with popular algorithms in references like Niblack and Threshold Image algorithm.Fourthly, our feature extraction and recognition method are proposed. Structure features extracted form binary images, like integral histogram and Moment Invariants, are tested on our database. Further more, Local Binary Patterns (LBP) is extracted from gray images as a texture feature. And some improvements of LBP are studied for the hand-dorsa vein images. An identification rate of more than 98% is gained on a database of 2040 images using Partition LBP (PLBP).The frameworks of the whole hand-dorsa vein pattern identification system would be discussed in this dissertation. And the practicability of vein pattern recognition technique is verified. The objective of building a practical recognition system is accomplished from both of the hardware and software.