Abstract
The separation of clods and stones from tubers is one of the main challenges in potato harvesting. A novel image processing technique was used in the present research for the intelligent separation of potato tubers from clods and stones. Digital images of 600 samples including 400 potatoes, 100 clods, and 100 stones were individually captured using an image acquisition system. The optimum imaging condition was found by histogram analysis. After image preprocessing, colour and texture features were extracted from different colour spaces and a subset of the most effective features was selected for classification using linear and quadratic discriminant analysis and artificial neural networks methods. Two-way and three-way classification strategies were considered and the results showed that the novel image processing technique using artificial neural networks can be successfully used for separating potatoes from clods and stones with acceptable accuracy.
