Research Of Preprocessing Technology In Cheque Recognition System

After the completion of the national check phantom exchange system, more and more financial instruments are used today, this bring lots of workload and pressure to the bank. Therefore numbers of scholars have already started to study the automatic recognition of instrument system, and hoped to solve this problem. But in the automatic recognition of instrument system, the preprocessing of recognition system holds a very important role. The result of the preprocessing system has a direct bearing on the merits recognition system's performance, so the research of the preprocessing system has a very real sense.In this thesis, the preprocessing technology consists of four parts:First take the check correction for grade and the binary technology; Next removing borders of amount in figures for the financial instruments; smoothing the noise of images; finally the segmentation of Arabic numerals.This main research work of this thesis is focused on the correction of check, the removal of borders of amount in figures and the segmentation of Arabic numerals. After gaining a check, for some kind of reasons, the check are inclined, in this article the Hough transformation is selected to take the straight line examination method and correct the check. In the binary aspect, this article selects twice Threshold segment binary algorithm to change the check to binary image. The experiment proved that this binary method obtains a good effect. In the remove column of small amount, on the basis of a careful reading in the related documents, the predecessors of the proposed approach are summed up and improved, coming up with the prior information of removing the line and strategy of filling the lack strokes. Experiments have proved that this method can be applied to the application of our current financial instruments and be able to achieve certain effects. In the segmentation of Arabic numerals, on the base of water drop fall algorithm and with the prior information of border, this thesis makes an improved water drop fall algorithm. This method has been incorporated into complete bank check character recognition system. It shows promising results.This article first chapter introduces the research of financial instruments preprocessing technology and describes the difficulties and broad application prospects of the preprocessing system in the recognition system of financial instruments. The second chapter introduced the check tile correction and binary technology. The third chapter introduced the improved algorithm of removing borders of amount in figures. The fourth chapter introduced the smoothing noises algorithm of binary image, and making several new templates to smooth and decreasing noises. Finally, this thesis introduced the improved water drop fall algorithm to segment characters.