Published in Scientific Papers. Series B, Horticulture, Vol. LXIV, Issue 1
Written by Rodica SOBOLU, Mirela CORDEA, Ioana POP, Luisa ANDRONIE, Dana PUSTA
In this work we proposed an automatic sorting algorithm for potatoes based on computer vision techniques. We performed two types of sorting: one depending on the potatoes size and the other one depending on their quality. We proceeded with the segmentation of the defected areas through global thresholding methods. Then, we extracted some morphological and statistical features from the segmented areas. These features were chosen as inputs for classification algorithms. We trained the SVM, Tree and LDA classification learners implemented in MATLAB Classification Toolbox and evaluated their performance. We concluded that the SVM has classified the potatoes according to their size with a higher success rate. In the case of quality sorting, the LDA method is recommended.
[Read full article] [Citation]
