Research And Design For Near Infrared Hand Vein Recognition System

This dissertation mainly deals with the research on a kind of important biological feature identification technology--the hand vein recognition. including the near-infrared hand vein image acquisition; image preprocessing; hand vein enhancement based on adaptive filtering method; hand vein segmentation algorithm based on the dynamic global threshold; hand vein feature extraction; classification design and system implementation. The main research contents are shown as follows:The principle of near-infrared hand vein imaging. Experiments show that the vessel structure can be more clearly displayed when the back of the hand is shined by the wavelength range between0.72～1.10μm light, which belongs to the near infrared light.The methods of image preprocessing. the images acquired by the equipment will inevitably influenced by the noise caused by the device itself or outside light. so preprocessing the images to remove noise is very necessary before the image recognition. What we are interested in is hand back vein information; however, the images acquired by the equipment not only include the hand back vein. but also include the skin of the dorsum of hand, so it is necessary to extract the interested region of he images, i.e., to extract the part containing the main vein information from the images; Moreover, because of the interference caused by the different skin colors and the external light, the collected images have large differences in gray. In order to facilitate later, we need to make gray-scale normalized for the images, i.e., to make all of the images have uniform intensity and variance.Hand vein enhancement based on the adaptive filtering method. Based on the Retinex theory, the acquisition of the camera’s image color or gray level is determined by the reflection light of an object, and has nothing to do with the intensity of reflection light; the color or gray of the object has nothing to do with the light irradiation uniformity. So the image adaptive filtering enhancement algorithm is to eliminate the interference of illumination component from an image, so as to obtain the reflection component, namely the essential features of objects. Adaptive threshold segmentation method. Although in the image after adaptive filtering, the hand vein has been significantly enhanced, at the same time, some noises are also introduced to the skin of the dorsum of hand. We can use the adaptive threshold segmentation method to remove the noise interference of hand skin, resulting in venous structure more clearly.A feature extraction method based on wavelet decomposition and mean absolute deviation. We analyzed the sub-band images of the hand vein image after the wavelet decomposition, and found the advantages of the low-frequency sub-band image as the hand vein recognition feature which can greatly reduce the image dimension. We finally present a feature extraction method based on wavelet decomposition and average absolute deviation, and the feature vector with48dimensional is generated. The Euclidean distance classification algorithm is used to classify the images. The experimental results show that the proposed method is not sensitive to small shift, rotation and scale changes. The total recognition rate is98.36%and the false acceptance rate is0.Design and implementation of the hand vein recognition system. The hand vein recognition system is divided into several parts:learning, testing and identification. In the learning part, the specified hand vein images are used to train the system, and the learning results are stored as a reference database for later testing and identification; In the testing part, the recognition correctness of the hand vein recognition system is tested; In the recognition part, a given vein image is judged that it belongs to which person in the database, and the system has the function of the refusal to identify. Finally, in the MATLAB7.6.0environment, a hand vein authentication system was designed and tested, the experimental results showed that for each hand vein image, the total authentication time is less than500ms, which meets the requirements of practical applications.