Tomato Classification and Sorting with machine vision using SVM, MLP, and LVQ

ABSTRACT: In the current paper, automatic control of tomato quality is analyzed based on using three different methods: LVQ, MLP, and SVM. Images are first captured by a digital camera and then denoising and contrast improvement operations are performed on them. Subsequently, the main step i.e. extraction of tomato features is carried out. The extracted features include: degree of redness and yellowness obtained in fuzzy form, greenness degree, first moment, second moment, third moment, average of these three moments, roundness value, and surface area. The obtained features are given to three different classifiers and the final results are compared and evaluated. The results suggest that SVM has a better performance compared to two alternative methods.