The Detection Of Paper-money Based On Machine Vision

With the great development of economy, the circulation of commodity and money become faster and faster. At the same time, there are more and more damaged and defiled bills. These bills result in a inconvenience of people' daily life which become a block of the commodity circulation. And bills used in ATM specially demand high quality. But in bank, the bills classification is still accomplished by manpower that is inefficient. The work of bills classification bocomes a burden in finance system. So the domestic demand of bill sorter is very exigent. The existent bill sorters are faulty and mainly import. For consummation of indigenous bill sorters, on basis of reviewing the existent effort, the author brought forward arithmetics of detection of damaged bills and classification of new and used bills which is the main function of bill sorter after investigation with the digital image processing technology. And the arithmetics are validated in emulator.Firstly , developmental current of the bill sorter is summarized, and the methods of investigation are compared. It is a important field of using digital image processing technology in the detection of paper money based on machine vision.Elementary knowledge and methods of digital image processing are then introduced. Histogram and color model of it are emphasized. After analyzing process of pattern recognition and composing of the system, the theory of statistical pattern recognition, fuzzy pattern recognition and frame-based pattern recognition are briefly described.With VC++ and MATLAB, author analyses the bill's image character specially. A detection arithmetic of damaged section is then brought forward which is different from optical and mechanical measures reported. Types of damage of bills can be detected by the arithmetic that the means of histogram threshold is adopted. Two innovative arithmetic of classification of new and used bills are introduced. The bills can be identified in the first one by calculating similar quotiety of histogram of bill's images in which difference exist. The histogram can be departed to calculate similarity. In this way, types of defiled bills are classified, and speed of calculation isimproved at same time. The second arithmetic results from disposal of bills' color images in HSI color model. The difference of H histogram and S histogram of the images can be used to classify new and used bills. The arithmetic excludes the effect of illumination in captured images, which is the central characteristic of the arithmetic. With the calculation of designed similar quotiety, validity of the arithmetic are proved by emulator in the space of small sample. The arithmetic introduced in this paper is innovative.At last, maturely applied arithmetic at present of value recognition, orientation recognition are briefly described. A soft flow chart of function and a brief chart of system of bill sorter adopted digital image processing technology and DSP are introduced.