RECOGNITION OF THE NUMBER AND HOLOGRAM

In the image analysis process, one of the most important steps is the segmentation, where interesting objects in the image can be identified. Without correct segmentation these objects can not be recognized. In this project segmentation is quite easy, because after correct banknote localisation and rotation the position of the banknote is static and well known. In this way it is quite easy to separate interesting region and process image just in some small area. This makes calculations a lot of faster and it is perfect way of optimization.

To recognise the nominal of the banknote the program generates histogram patterns (vertical and horizontal) and looks for correct pattern option for every case. By histogram patterns of nominal value is easy to recognise 5 € and 10 €, 20 € and 50 € are very similar and for their distinction are used other algorithms like hologram analysis (for distinguish 50 Euros) and colour analysis. By algorithm of number recognition it is also easy to say if some banknote is corrupted. If the banknote is corrupted or has bad position, bad quality or is another object, then the calculations are finished and program returns zero result.

Another option for recognising the nominal is to analyse the hologram. For 5, 10 and 20 Euros the hologram on the right side is in form of line, for 50 Euros it is different and has form of small quadrant. This method is good to recognise the nominal of 50 Euros. The biggest problem of hologram recognition is that is very sensitive for light and usually is distorted by light reflexes. However, the method is quite easy and simple and it is used in the program as a part of recognition algorithms.

Hologram analysis

Every Euro banknote has hologram. On the front, the hologram is in the form of a stripe for the 5, 10 and 20 Euro banknotes and a patch for the 50, 100, 200 and 500 Euro banknotes[8]. Because the program does not recognise 100, 200 and 500 Euros, the patch hologram would say with high probability that the processed banknote is of 50 €.

To perform this task, the program looks for interesting region, and then with the help of histograms, examines if the histogram fits the pattern. For this small area the program calculates horizontal histogram (sums of columns from 0 to width).

To recognise the nominal the program uses the patterns. A pattern is a type of theme of recurring objects, in case of Euro recognition program a pattern is a part of image, where generated horizontal and/or vertical histograms have some determined properties. For example, the pattern for 5 Euros has first non-zero area (number 5) bigger than second non-zero area (small Euro letter) in horizontal histogram. The pattern for 10 Euros has first nonzero (that is number 1) area smaller than second non-zero area (number 0). Pattern for 20 and 50 Euros are very similar and first and second nonzero areas have similar width.