Font Size: a A A

Studies On Seal Identification System

Posted on:2009-12-26Degree:MasterType:Thesis
Country:ChinaCandidate:H Y SunFull Text:PDF
GTID:2178360242974973Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Although biometrics has been used in many areas, seal imprint as a kind of personal identification is widely used in Japan and some other Asian countries. Seal imprint is used in many types of text and most of seal identified by human eyes. With the popularity of information technology and office automation, the traditional means of seal management has been unable to meet modern needs, especially in the public security and financial fields that need to use information technology to identify more urgently. Therefore, this paper introduce the demand in home and the current research abroad of seal imprint, and then particularly discuss identification such as the extraction of seal color, binary, image denoising, at the end of the paper we introduce the identification based on texture features. As follows:(1) Analyses the cheque's current research and development and then study the cheque's characteristics and the development direction at home. This paper pointed out the existing problems of cheque identification and the purpose and significance of topics;(2)In the part of preprocess we studied seal segmentation, binary and seal denoising based on HSI color space;(3)Base on the two seals' deflexion angle gained in parameters calculating to rotation matching after making translation transform the two heart-shaped coincidence;(4)Put forward an identification method based on the texture of the spectrum and SVM, this method improved the other methods' identification error caused by rotation, effectively solved series difficulties in practical application such as the influence due to chop forced uneven of human factors and the quality of ink.The experimental results show that, the identification method based on the text in the texture and SVM can effectively classify the different types of objects in characteristics space and can classification correctly in the reasonless seal images, improve the recognition rate.
Keywords/Search Tags:segmentation, registration, texture, support vector machine
PDF Full Text Request
Related items