Money Money TEAM 6 The TEAM Dana Damian

Money, Money TEAM 6

The TEAM Dana Damian Scientist Institute: Politehnica University of Timisoara Country: Romania Krisztina Dombi Documenter Institute: University of Szeged Country: Hungary Levente Sajó Programmer Institute: University of Debrecen Country: Hungary Zoltán Horváth Institute: Pannon University Country: Hungary Gopher

The Problem: Counting money. Input: Photo of coins (EuroCent perspective view, non-uniform lighting, eventual partial covering) Task: Recognize the coins and count the total sum. Output: The sum, and also the recognition statistics (accuracy / false positive rate etc) of the implemented method. Difficulty: Medium

Our Problem The Problem: Counting money. Input: Photo of coins (forint with perspective view, without covering) (Let’s say we have a lot…) Task: Recognize the coins and count the total sum. Output: The sum, and also the recognition statistics (accuracy / false positive rate etc) of the implemented method.

Motivation • In business transactions, to enable computers to recognize coins and other different forms of currency has become an essential process. • If computers are able to do the recognition, all monetary trades and transactions will be much easier. • Our scope is limited on recognizing only the Hungarian coins ( head OR tail ) (1 F, 2 F, 10 F, 20 F, 50 F, 100 F).

Monetary automates

Handy coin counter

Approach • The application is suitable for an architecture of a coin counter system that incorporates a steady camera which monitories coins passing beneath (maybe on a belt )

Catalogue of Hungarian denomination

Theoretical background of Hough transformation • A transformation that maps a point in a Cartesian space onto a 2 D space of points, called the Hough Space

Circular HT • Extension of the classical HT • Analytical function of a circle leads to a mapping of each point (x, y) from the image onto a 3 D Hough Space parameterized according to (a, b, r) tuple, where – (a, b) center of the circle – r radius of the center Points satisfying the equation are mapped into the accumulator according to the circle they belong to

Preprocessing Enhance Contrast Sharpen Gaussian Blur Sharpen Find Edges Threshold Fill Holes Outline Invert

Hungarian coin counter system Input image:

Enhance Contrast

Sharpen

Gaussian Blur

Edge Detector

Threshold

Fill Holes

Outline

Invert binary

Circular Hough Transform

Detected coins

Center points and radius

Result

Core Idea • Having a picture for training purposes, the system designs a coin table in which it stores the size of each coin • Further recognition is based on comparison with the coin table

Main issues • Shadows can enlarge the image of a coin, thus increasing its radius • Different condition of illumination can generate an edge map with lack of information • Coins are very close to each other

Limitations • A priori knowledge of the # coins • Dependence on the quality of edge detector

Future Plans • Go to the Bajor söröző • Eat good and drink a lot • Go back to the dormitory • Go home with lots of new experiences, new remembrance

Other Works • • • Coin Detector CS 7495/4495 Term Project Dong-Shin Kim(gtg 901 p) CS 7495 Young Gyun Yun(gte 257 z) CS 4495 You-Kyung Cha(gte 440 y) CS 4495 Dagobert – A New Coin Recognition and Sorting System Michael N¨olle 1, Harald Penz 2, Michael Rubik 2, Konrad Mayer 2, Igor Holl¨ander 2, Reinhard Granec 2 ARC Seibersdorf research Gmb. H 1 Video- and Safety Technology , 2 High Performance Image Processing A-2444 Seibersdorf Design and Evaluation of Neural Networks for Coin Recognition by Using GA and SA Yasue Mitsukura*, Minoru Fukumi* and Norio Akamatsu* * Department of Information Science & Intelligent Systems, Faculty of Engineering University of Tokushima 2 -1, Minami-josanjima, Tokushima, 770 -8506 JAPAN

Thanks for your attention. Questions? …