RECOGNITION BY THE COLOUR

Colour analysis is very difficult task, because colour can change on every picture depend on light. Colour is the product of a visual environment, illumination and the human brain. The perceived colours of objects are determined by the spectrum reflectance of surfaces, illumination, and the visual angle; illumination and the visual angle are always in flux.[9]. In Euro recognition program colour recognition algorithm is used just for nominals of 20Euro and 50 Euro. It is because algorithm of nominal recognition give good results for 5 Euro and 10 Euro and it is not necessary to use other additional algorithms. For 20 Euro and 50 Euro nominal analysis does not give correct result and it is necessary implements additional method just for those two nominals. The colour recognition is simplified to the problem of deciding if the analysed colour is more blue or more red.

In the presented algorithm first step is to extract a small region of interest. Sometimes the picture of banknote can move a little bit depend on the quality and rotation of the photo, so the program just check if the picture is more red than blue or more blue than red.

Some face recognition algorithms uses the property of YCbCr colorspace that pixels belonging to skin region exhibit similar Cb and Cr values. 50 Euros has a colour similar to human skin colour and the program for Euro recognition use this characteristic.

On the other hand very good results gives Hue value analysis from HSV colorspace. In the program pixel with H-value between 150 and 200 is classified as a Blue and pixel with H-value between 0 and 20 is classified as a Red. The banknote is recognise depend on which value is grater: if blue then recognise 20 Euro and if grater is red then program returns 50 Euro. For 10 Euro and 5 Euro colour analysis is not necessary because in theory the program should recognise them in nominal analysis algorithm phase.