Computer vision based currency classification system

Abstract

There are numerous problems associated with current system which are solving the problem of automatic currency classification. Some of the problems administered are like scaling, rotation and noise in the form of missing valuable data in printing or due to the wear and tear of currency notes. In our system we are first aligning the image horizontally along the x axis and after that foreground of the image is removed by detecting the location of edges, and once we have got the processed image we can apply any of the techniques for classification. Over here we are using fast template matching for recognizing the value of the currency. Once we get result after template matching we can classify the currency into different categories like 10, 50,100,500 and 1000. In our system we are aiming at the improvement on existing system by adding useful and robust pre-processing techniques which has been missing in most of the recent works done so far.

The State-of-the-Art Technology of Currency Identification

“…One generally used in feature extraction is Canny edge detector. There are two criteria for evaluating performance of edge detectors [1] [38] [49]. The first criterion is that a qualified detector needs to have a good signal-to-noise ratio which is able to facilitate the process of edge detector.…”

Currency security and forensics: a survey

“…A modified Canny edge detection is used [62] for removal of noise and background from captured images. The Canny algorithm marks a point as an edge if the amplitude is larger than its neighbours without checking that the differences are higher than expected.…”