Algorithm Research For The Non-Touching Banknotes Counting System

Currently, in banknote producing industry, what used in the banknotes number management is small size counter, which is mechanical and has a lot of defects such as noisy and may damage the banknotes. A non-touching banknotes counting system based on the image processing technology can overcome these defects. It get the number of the banknotes by taking photos of a side of a bundle of banknotes by a special camera, and dealing with the picture with some special image processing methods. The difficult of the design of this system is algorithm design.Nowadays, the algorithm of the non-touching counting system has four steps. First, segment the picture with an adaptive threshold calculated before and get the binary image. Second, search the striation lines in the binary image and mark them. Third, make the striation lines thinner and get rid of the noise. Finally, calculate the number of the striation lines and get the number of the banknotes. There are many defects in the algorithm above. First, a segmentation method based on threshold can't abstract the texture from the noisy and low-contrast original picture. Secondly, the texture thinning after searching is needless.According to the above-mentioned defects, this thesis offers a new non-touching banknotes counting system algorithm, which includes a texture abstracting algorithm based on gray compare, smooth in space domain, Gabor filter, orient filter and segmentation by threshold and a texture searching and counting algorithm based on continuity analysis. Experiments proved than the veracity of it can reach above 95 percent.