Preliminary study on color based nectarine variety classification

This work presents a preliminary study on an automatic color based nectarine variety classification system. This system is planned to operate in a fruit packing line in order to contribute to increase the quality of an industrial post-harvesting process by verifying nectarine variety. This proposal is based on the computation of the histogram of the hue, saturation, and value (HSV) skin color description of the nectarines analyzed. The histogram is operated as a fingerprint and used to classify the nectarine varieties by defining a reference database with target varieties. Six different varieties have been analyzed in this preliminary study: one yellowish and five reddish. Results show a classification success of 70% for the yellowish variety whereas two reddish varieties show a very low classification success, 20%. These preliminary results evidence the complexity of the classification problem due to the skin color similarity between varieties and the associated quality production problem as different varieties may differ in taste and/or flesh color.