INSPECTION OF TOMATOES USING IMAGE PROCESSING TECHNIQUES

Abstract

This work aims to inspect tomato features and classify them based on color and morphological features into the three predefined regions using artificial neural networks (ANN). Different learning methods were analyzed for the task of inspecting tomatoes using image processing software in MATLAB. Tomatoes were collected from the eastern parts of Ethiopia. The neural classification was done by the shape and size feature alone. The ANN classifier on the selected color feature alone showed that from the total test examples of 180 images, 168 (93.3) were correctly classified and 12 (6.7 %) were misclassified. The ANN classifier on all features taken together showed that all the test images were correctly classified. This result is similar to the morphology (shape and size) features result, but if the number of data points is high, the result may vary significantly. The overall result revealed that shape and size features have more discriminating power than color features, and the discrimination power increases when individual features are trained together with shape and size features. This may be because the discriminating factor increases due to the increase in the number of included features. It was observed that the proposed method was successful as quantified by the cumulative error (CE) and percentage error (%E) of training, testing, and validation of color features: 6.35 %, 3.70 %, and 11.11 %, respectively, in evaluating the quality of tomatoes.