Fruit Classification by Extracting Color Chromaticity, Shape and Texture Features: Towards an Application for Supermarkets

In this paper we present an approach for fruit recognition using artificial vision, towards to employ it as an application for supermarkets. The fruit's visual features extracted are shape, texture and -we focus on- the extraction of the color chromaticity. From the few related works on fruit recognition, the color extraction is performed in the RGB (Red, Green, Blue) space. We claim as necessary to employ the chromaticity of colors to characterize the color of the fruits. Thus, we propose to use the HSV (Hue, Saturation, Value) space because it is possible to extract and process the chromaticity data, but without the undesirable intensity effects of the RGB space. On the other hand, before the color is characterized, we complete a selection of the chromaticities that contribute with important data about the fruits. Our approach's strength is validated by performing test on 20 common fruit classes: this proposal outperforms the color characterization of related work.