The Synthetic Analysis Methods of Computer Seal Identification

摘要

The thesis presents the synthetic analysis methods of computer seal identification. To identify seal, first, this paper presents a new method of threshold selection for binary images. It is based on only a simple image statistic and is easy to compute and effective. Second, the feasibility of computer seal identification is discussed and some useful features and identification methods are proposed. According to the characteristics of seal image, here, we propose some translation-invariant and rotation-invariant features. They are Gross-mass, Seal-mass, Distance between centers and Maximum correlative value of the Ring-Dates. Our experiment shows that these features are easy to obtain and powerful. Finally, the process computing model value is discussed in details. The results of experiments are given.