Iranian Research Organization for Science and Technology (IROST) Innovative Food Technologies 2783-350X 5 1 2017 11 22 Development and evaluation of artificial neural models for RGB to L*a*b* color feature transformation using machine vision system Development and evaluation of artificial neural models for RGB to L*a*b* color feature transformation using machine vision system 151 158 469 10.22104/jift.2017.469 FA Sajad Kiani Ph.D student of Biosystems Engineering Department. Tarbiat Modares University, Tehran, Iran Saeid Minaei Professor of Biosystems Engineering Department. Tarbiat Modares University, Tehran, Iran Journal Article 2016 12 15 In this study an intelligent system based on machine vision, Multilayer Perceptron (MLP) artificial neural network and Partial Least Square (PLS) models was developed to estimate the L*, a*, and b* values for saffron samples utilizing their RGB color values. Color images of 33 saffron samples (165 images) and 150 color images of standard colored plates were captured utilizing the developed machine vision system. In order to extract RGB parameters, the images were processed using image processing algorithms. Also, L*a*b* values of each sample was measured using a commercial colorimeter (Hunter Lab, color Flex, USA) in triplicate and the measurements were averaged to obtain the final values. RGB values and their linear transformations were set as the inputs of the models and L*, a*, and b* values were set as model outputs, respectively. Experimental results showed that the performance of MLP models were better than those of PLS, with high correlation coefficients of cross validation (R² and RMSE values equal to 99% and 0.769, 0.953, and 0.785, respectively). Finally, it can be stated the capability of machine vision technology for color quality evaluation of saffron. In this study an intelligent system based on machine vision, Multilayer Perceptron (MLP) artificial neural network and Partial Least Square (PLS) models was developed to estimate the L*, a*, and b* values for saffron samples utilizing their RGB color values. Color images of 33 saffron samples (165 images) and 150 color images of standard colored plates were captured utilizing the developed machine vision system. In order to extract RGB parameters, the images were processed using image processing algorithms. Also, L*a*b* values of each sample was measured using a commercial colorimeter (Hunter Lab, color Flex, USA) in triplicate and the measurements were averaged to obtain the final values. RGB values and their linear transformations were set as the inputs of the models and L*, a*, and b* values were set as model outputs, respectively. Experimental results showed that the performance of MLP models were better than those of PLS, with high correlation coefficients of cross validation (R² and RMSE values equal to 99% and 0.769, 0.953, and 0.785, respectively). Finally, it can be stated the capability of machine vision technology for color quality evaluation of saffron. https://jift.irost.ir/article_469_dc6a6489640ca02b0d42dabeb8e46bb7.pdf