The Design Of Intelligent Cash Counter Based On Image Recognition

Recent years, with the high speed development of economics in our country, commercial behaviors are happening every minute. As a result the population of currency becomes bigger and bigger. Therefore, the cash counter plays a more and more important role in dealing with daily operations in bank. Presently, in market, most cash counters use the size of a note to recognize its values. But the disadvantage is that to various versions of the same notes the error rates are very high, so various types of notes can not be counted at the same time.The common counter has ability of dealing with notes of same versions and checking if it is a fake based on "magnetism", "ultraviolet light" and "infrared light" techniques. In this dissertation, a design method of an intelligent cash counter is put forward based on the image processing technology. Using this method, the intelligent counter can efficiently recognize the values of notes, in various versions and strange paper notes. So the function of current counter is changed from "counting numbers" to "counting money", and the digitization, intelligentization and humanization is realized in the design of cash counter.This system focuses on sampling the paper note image in high speed, pre-treating the image data and recognizing them within the duration of paper note's going through the counter. Then the result of recognizing is conveyed into a controller. The difficulties in this process are the high-speed sampling, processing and recognizing of image. The algorithm of processing and recognizing of the paper note image is studied in details in this article.The traditional methods of processing and recognizing of image are discussed. The image pre-treating is dependent on five operations, sampling, edge detecting, declining adjusting, data binarizing and noise filtering. But these common methodsare not suitable for this system because of its limitation of real-time. So these methods are improved in the dissertation and a new algorithm, partial features detecting method, is put forward. And an effective recognizing method based on template confirmation is introduced too. So the real-time capability and recognizing effect are achieved.