Seal Imprint Recognition Based On Improved Artificial Bee Colony Algorithm

In banking business,as one of essential recording items,affixing the seal imprint on the notes is the behavior of assuming notes liability and protected by law.For this reason,the seal imprint identification has become a important work which bank staff face everyday.However,the traditional method of artificial seal identification relies on operators’ eyes and work experience,so it exists malpractices of slow computation speed and low accuracy or other reasons in image matching and doesn’t adapt to the rapid development of economy and society.In the same time,the forging methods appear one afer another along with technological improvement of making seal.As a result,artificial methods are difficult to identify the differences of true and forged seal imprint,becauseare they’re very seldom often.A new technology is needed to solve the problem in these cases.The development of image processing technology and pattern recognition provide a new approach to this kind of problem.The research object of this paper is seal imprint on notes.The paper proposes a method based on improved artificial bee colony algorithm of seal imprint identification.At first,preprocessing method of the seal imprint image.From the RGB color model,the method changes color image to gray image by differences between red component and other components,and sets a threshold for image binarization.Then,it analyses seal imprint interference factors and removes writing lines and characters separately,and applies median filtering to eliminate noise in the image.Secondly,the seal imprint identification.The paper establishes a uniform image coordinate system,and uses improved artificial bee colony algorithm which takes defined dissimilarity function as standard,and renews probability formula,and next searches for movement values and rotation angles which demand matching of reserve seal imprints and unidentified seal imprints.Finally,the paper distinguishes the true seal from false ones by matching reserve seal imprints and true seal imprints or forged seal imprints separately.By experiments,it determines a dissimilarity threshold to measure true or forged seal imprints,and achieves seal imprints identification.The experimental results show the method which the paper proposes has advantages of good applicability and high precision,so it is applied to the seal imprint identification working practice.