Classification of bitter and sweet olives using image processing and artificial neural networks during curing process in brine and water environments

Abstract

The external appearance of an olive’s skin is the most decisive factor in determining its quality. In order to quantify the colour changes that occur during sweetening process, a computer vision systems based on Artificial Neural Networks (ANN) was developed to determine bitter and sweet olives. To this end, olives were treated in brine and water testing environments for nine days. During this process, digital colour image were taken every two days. Two colour spaces (RGB and L*a*b*) were used and six colour features corresponding to average of each colour channel were derived and used for ANN training. The overall accuracy of ANN classifier was of 93.38%. According to the results of the image analysis, the L*, R, G and B values showed a significant correlation with general acceptance for water curing treatment (p < 0.05). For brine curing treatment, results showed a significant correlation between L*, a*, b*, R and G with general acceptance (p < 0.05). Considering the results of sensory analysis the optimum time for both brine and distilled water were seven and three, respectively. Furthermore, brine curing had a greater impact on olive skin-colour changes.