关键词: 手背静脉识别, 图像定位, 有效区域, 二维Fisher线性判别, 最近邻分类器

Abstract: The palm-dorsa vein recognition uses non-contact technology. A locating algorithm was presented in order to avoid the interference brought by the rotation and translation of the palm-dorsa when collecting pictures. This method ultimately intercepted a rectangle on the palm-dorsa containing the information of vein as much as it can. The feature points of palm-dorsa edge were located by the location method based on unchanged feature points, and then the effective region was obtained. Experimental results show that the algorithm is self-adaptive, accurate and fast. To solve the problem that the within-class scatter matrix is always singular based on Fisher Linear Discriminant (FLD), a new way based on Two-Dimensional FLD (2DFLD) was used in the palm-dorsa vein recognition. The image matrix was directly projected to avoid high dimension operation. In the experiments, 2DFLD was applied to extract the vein feature subspace from palm-dorsa vein database. The test sample images to be recognized were projected on small dimension subspace. Lastly, a nearest neighbor classifier for palm-dorsa vein matching based on Euclidean distance was used and the recognition rate reached as high as 98%.

Key words: palm-dorsa vein, image locating, effective region, Two-Dimensional Fisher Linear Discriminant (2DFLD), nearest-neighbor classifier