Extractoin Of Eigenvalue On Fiber Lines

In recent years, fiber in information security field is becoming a new material.We use thermal characteristics, photosensitive properties and geometric characteristics to identify.Based on the analysis of domestic and international anti-counterfeiting technology of the fiber, we studied the characteristics of the image. We can extract the geometric feature's parameters of the fiber image we captured. The parameters are unique for each fiber picture, so we can bind the parameters as unique source correspondingly. This thesis about image recognition algorithm focuses on the following aspects:First, we focus on preparation of the paper embedded security fibers, and collection of original gray-scale digital image.Second, we focus on the security fibers'image pre-processing: we certified that fibers'images denoising are necessary. In this paper, we use the square root of the wavelet denoising algorithm for image pre-processing. We certified that the algorithm in the fiber image denoising performances excellent. We can smooth region without destroying detail information.Third, we focus on image segmentation.We compare local uniform measure method with threshold image segmentation method.Fourth, we focus on image thinning. We design thinning algorithm combining classic central thinning algorithm method and Central approximation algorithm.Last of all, we focus on feature extraction and matching. We propose that angle of the lines vector can be fiber's charactors.