Automatic Classification of Objects on their Shades of Colour using Consumer CMOS Sensors (WhiteStone)

The main goal of the WhiteStone Central Innovation (ZIM) project, which was carried out with an industry partner, was the development of a camera-based unit for the correct identification of the colour (Yxy value) of objects. The difficulty in doing so lay in the changing colour temperatures of the lighting, as well as in any reflection which occurs. Nevertheless, the smallest differences in colour were to be reliably recognized. For this an algorithm was developed which carries out a correction of the colour temperature in order to obtain absolute colour values. During the project an LED-based light source was developed which can control the colour temperature almost at will through the use of a true-colour colour sensor. For the classification of the colours a neuronal network was conceived, which was able to ascertain the colours very reliably.

The photograph show the histogram of two objects which can only be differentiated from each other with great difficulty by the human eye. However the neuronal network is able to segment them reliably.