Estimation of tomato [Lycopersicon esculentum] ripening stages using three color models

This research was carried out to compare the effectiveness of three color models in estimating the ripening of tomato fruits using features from image histograms and linear discriminant analysis as the statistical classification model. Digital color images were taken from nine tomatoes for each five maturity classes for a total of 45 samples. In each class, five samples were used for model development while the remaining four samples for model verification. Using Matlab (version 6.0 Release 13 Image Processing Toolbox ) , the images were processed to compute their histograms using the RGB, HSV and CIE L*a*b* color models at different bin sizes. Linear discriminant analysis using a statistical analysis software (SPSS) was performed on the histogram features to determine a multi-variate classification model. While all the color models had successfully classified 80 to 100% of the model development samples, they did not performed so well in the verification samples. The average classification success rates of HSV (62.5 %) and CIE L*a*b* (60.0 %) were almost the same and much better than in RGB (35.0 %). Increasing the number of bins did not, however, result to a better classification model.