Non-contact Based Palmprint Authentication

In recent years, Palmprint recognition emerging as a method of identification is an important complement to existing biometric technology. So far, most studies have focused on palmprint identification on palmprint recognition algorithm, one of the non-contact palmprint identification system consideration is how the hand is detected and segmented from the background. Non-contact acquisition mode imaging environment brought instability, palm position change increases the difficulty of locating the palm area of interest, especially the light of some changes have a serious impact on the ability to identify the system, In the practical point of view, how to improve the non-contact palmprint identification system under variable illumination and complex background environment which is robust, to produce the product to the user friendliness and security sex are our attention.This paper studies how to improve the success rate of hand collected in a non-contact condition,Summarized as follows:1. A study of Haar features of hand. Haar features from the form typicall comprises two or three coupled black and white rectangles from the performance value of black and white rectangles rectangular sub-pixel gray values in the image region corresponding to the windows for the sum of the difference that can reflect changes in the local gray image clearly. e.g., human eye portion with respect to the face clip part of the face, these eye features can be characterized by a very good performance with a prototype; For the hand, in the background there is cross of your fingers and will also appear in the image area showing different characteristics.2. A set of hand image samples on hand to outline a number of feature points to describe the whole hand, then use the principle of matching analysis of these models, to calculate the average model, scaled to fit the hand, flip characteristics.3. Improved YCrCb space algorithm. The traditional color space has obvious short-comings, for the color difference signals Cr, Cb values,they have had good together feature under intense lighting conditions class resistance, after adjustment of traditional spatial transformation algorithm to study.4. The traditional hand contour extraction methods are mostly based on skin color, has been greatly affected in a complex environment, how to extract the hand contour images precisionly. Here we propose two solutions:one based on the first palm model, Haar feature recognition to frame the image in the palm of the approximate location, and then we take a dynamic frame learning method to calculate accurately the palm outline; the second is when the hand image acquisition accounted for a larger proportion of the whole picture, the direct application of the regional dynamic learning to gain ROI.5.Wavelet decomposition palmprint recognition algorithm application were studied. The first is the statistical analysis PCA algorithm,then try to improve the high frequency recognition performance.6.Palm Images database under normal environement is founded which brings great benifit to the next experiment.