Hand Shape Verification Of Light Background Based On Shape Context

With the development of science and technology and expanding of human activities'range, our demand for security has been proved to be a rising trend year by year, biometrics has been widely researched and focused on for this reason. And the identification methods that could be used easily and calculated simply got more and more attention of people with the development and mature of technology. The proposed hand shape verification method in this paper is a biometric identification technology which is easy to be accepted by the user for the advantages of easy to implement and low cost.The existing hand shape verification method often sets the background to black when acquiring the images; therefore most of the scholars only focus on the extracting and matching of hand shape feature instead of hand shape contour in the process of certification.With the deepening of the research, fixed black background has couldn't satisfy the needs of reality for its limitation on the application environment of hand shape verification. The study of hand shape verification methods in this paper is made under the light background of image acquisition, and the limitation on the user and the environment of hand shape verification has been reduced by increasing the applicability of the method.The hand shape image that acquired under light background is vulnerable to shadow,which will bring great difficulty to hand shape contour extraction. In this article, the complete hand shape contour could be acquired by image segmenting based on hand sample characteristics. First, segmenting the image and the point coordinates of the hand shape sample can be extracted. Secondly, the image was converted to the color space of YCbCr, and extracting the hand shape sample area according to the coordinate of sample points in the Cb,Cr component, then calculate the Euclidean distance between all pixels in the Cb, Cr component and sample characteristics, and saved them in a distance matrix, the hand shape outline can be extracted by threshold segmenting based on distance. Finally, recognition of the hand shape position, all hand shape direction would be unified by rotation. This method can extract the hand shape outline from the image of shadow background, which can be used for light and black background, and reduces the limitation on the user and the environment.In this paper, the hand shape verification was achieved based on the context, and a local shape context algorithm was proposed. First, intercept relatively stable outline from hand shape image, and achieving the sampling point set by local uniform sampling on the intercepting outline, then statistics the hand shape context features within the local scope,finally, matching the hand shape according to the extracted hand shape features, and identify the user by the matching cost.The proposed method of hand shape verification is robust and insensitive to the change of light conditions and background, the contour extraction method can be widely used for its good portability; The local context algorithm has improved the differences of hand shape characteristics between different users without increasing the same user hand shape matching cost. We have achieved the Equal Error Rate of 3.15% for verification. Compared with the other shape context algorithm, the Equal Error Rate fell by 1.5%-5%. The proposed method has a higher accuracy and is more suitable for hand certification.