Research Of Hand Vein Recognition Based On Gabor Multi-orientation Fusion And Multi-scale HOG Features

With the rapid development of information technology,people attach high degree of concern and continuous requirements to modern information security,identity authentication using individual biometrics has caused more attention.Biometric technology is an accurate and rapid identification technology that combines the characteristics of the human body with its own biological characteristics(such as fingerprint,iris,expression,etc.)and behavioral characteristics(such as handwriting,voiceprint,walking morphology)with modern information means such as optical,electrical,image processing and computer applications.Dorsal hand vein recognition is a new kind of biometrics technology for identification that rises in recent years.It obtains the hand vein distribution images of the individual through a specific acquisition system,and then uses certain image processing method and feature extraction algorithm to obtain vein characteristic values for the final feature matching.Compared with the traditional identification methods(such as passwords,keys,RF cards,etc.)and other biometric identification methods,hand vein recognition technology has more advantages,it uses live and no contact certification so that couldn't been lost ant no easy to forge,also it will not be subject to external environment(temperature and humidity).So hand vein recognition technology becoming the research hotspot of current biometric technology.In this paper,characteristics of dorsal hand vein are seen as the research object for recognition technology.We have made scientific research and summary of several components: collection of images,preprocessing,feature extracting and recognition.Besides,a new method based on Gabor multi-direction fusion and multi-scale histogram of Oriented Gradient(MS-HOG)is proposed which could improve the accuracy of hand vein recognition.The main work of this paper includes:1.The dorsal hand vein collection system has been improved for easy operation and obtaining images with better quality.2.A new hand vein feature extraction algorithm is proposed.Based on the multi-scale and multi-directional Gabor feature map,images in different directions of the same scale will be fused.Then,HOG with certain scale will be used for correspondent fusion graphs.Finally,HOG features of different scales would been connected for recognition.Experimental results show the effectiveness of the proposed algorithm.