A portable computer-vision-based expert system for saffron color quality characterization

Abstract In this work, attempts were made in order to develop and evaluate a Computer Vision System (CVS) for non-destructive characterization of saffron ( Crocus sativus L.). Thirty-three saffron samples from different geographical regions were tested. Fourteen color features were extracted using image analysis. Principal Component Analysis (PCA) was used for saffron sample clustering and for selection of color features. Partial Least Squares (PLS), Multiple Linear Regression (MLR) and Multilayer Perceptron (MLP) neural networks were utilized to establish relationships between color features and coloring strength of saffron based on ISO 3632 standard. Experimental results showed that the optimal PCA was obtained by the first 2 PCs and with 95% total variance between the samples tested. Performance of MLP models for saffron color characterization were better than others, with high correlation coefficients of the cross validation (R 2 and RMSE values equal to 99% and 4.5, respectively) and high classification success rate of 96.67%.