Research On The Algorithm Of Automatic Seal Identification System

With the development of social economy, the portfolio of the bank increases dayby day, so the traditional method of bank seal identification with the deficiencies of lowrapidity and low precision cannot readily meet the needs of banking transactions. Whilethere is only a single criterion in the existing bank seal automatic recognition system ofmarket with the deficiencies of low rapidity and low precision. Thus, carrying on furtherresearch of seal identification system algorithm of the bank has higher practicalmeaning and theoretical meaning. Ability to distinguish the forgery seal impressionaccurately and reliability of the differentiation of true and forgery, as well asadaptability of relatively great real seal difference, are characteristics of computermethod of automatic seal identification and difficulties lied in the research. On thefoundations of the research reports and scientific papers of the domestic andinternational seal identification in recent years, this paper has probed into and studiedabout each stage algorithm of the seal impression recognition system of the bank. manykinds of criteria combined together have been proposed, systematic pretreatmentstages have been paid attention to and improved to guarantee further real seal pass rateand system discern speed in case of near 100% of forgery discerning rate of seal. This text has launched the following work in algorithm research of each procedureof automatic recognition system of the seal impression mainly: The pretreatment improves the technology of cutting apart the seal impressionpicture based on color , which is transmitted from the cutting apart of the traditionalcolored picture based on RGB colour model to the seal picture based on HSI colourmodel. The paper has also proposed morphologic disposition of the cut seal picture inaccordance with the quality of the cut picture before and after the threshold. It proposes to discern seal type first and then to precision register withdifferent method in according to different type in the rotating registration, in order toenhance the precision of registration with possible influence of the registering stageupon the latter identification . It proposes to identify step by step at different stages according to multiplestatistic criteria with comprehensive identification in the use of seal imprint's basiccharacteristic, moment invariant and skeleton. According to the skeptical seal identification in the automatic seal II重庆大学硕士学位论文 英文摘要identification, it proposed further imitation of manual identification method so as toguarantee the nearly 100% forgery seal identification of the system.