Research On Palm Vein Extraction And Recognition Method

Because the palm vein is the subcutaneous tissue palm hand, its distributed information is difficult to be stolen. Recently the palm-vein recognition has become a new biometrical feature identification way. The subject's main work is that to direct at the prime functions in developing palm vein recognition system to do the research and design, to realize the models of capturing palm-vein image, preprocessing, extracting feature and matching and so on. And this subject mainly researches the present existing issues of location to the image of non-contact capturing device, the segmentation to the image having inhomogeneous illumination, low-contrast, severe noise, etc. and the recognition and matching.Firstly, the capturing method is summarized in this paper; Secondly, the location method is proved. In the course of capturing image, two images are obtained by the way of setting the device's parameters twice. The one picture is the white hand shape with the black background, the other one is palm-vein image. Using its advantage of feature easily on this kind of hand shape the Region of Interest(ROI) is located on it firstly, and then the ROI is extracted on the palm-vein image; And then direct at the poor segmentation quality of palm-vein image due to the problems of inhomogeneous illumination, low-contrast and severe noise in the palm-vein image, this paper proposes a method of the palm-vein image segmentation automatically based on local gray minimum. This method avoids thresholds selection and is able to solve effectively the disturbances of inhomogeneous illumination, low contrast, and severe noise. The result of the experiment that is taken in the palm-vein image database built by our laboratory proves that this method can restrain the affection that noise, inhomogeneous illumination and low contrast brings, in addition, it has the greater superiority than all the latest palm-vein segmentation methods in the respect of extracting the palm-vein line continuity and having no use for selecting thresholds, etc; At last, a matching method based on the structural feature of the palm-vein line is used. The accuracy and efficiency are improved. According to the methods above, the experiment is done through the image library built by myself including 583 pictures from 53 people and 11 pictures for one people under the condition of different times and different inhomogeneous illumination. The recognition rate is 97.94% and the run time is 4.114s. The results of the experiment express that the ROI location is realized in the non-contact capturing device; This segmentation method proposed can restrain the affection that noise, inhomogeneous illumination and low contrast brings, in addition, it has the greater superiority than all the latest palm-vein segmentation methods in the respect of extracting the palm-vein line continuity and having no use for selecting thresholds, etc; Contrast to the Point to Point, thick and thin,2DFisher and 2DPCA, the recognition rate is improved by this matching method proposed. The research purpose is realized.