Separating clods and stones from potato tubers based on color and shape

Abstract

The separation of clods and stones from the harvested potato tuber has always been a prevalent problem in the world. However, the precision of sorting was restricted by the potato tubers covered with mud on the surface. This paper studied methods of separating clods and stones from potato tubers based on shape and color. An image acquisition system consisted of a light source, a camera, a computer was built for this experiment. The color features were extracted from the components of RGB and HSV images by the two-dimensional Haar Wavelet Transform and put into SVM (support vector machine) to classify the object after principal component analysis. The shape features which contained the original contour and corrected contour described by the mathematical statistical methods was extracted and used for separation by SVM. The experimental result showed that it was effective to separate clods and stones from potato tubers based on the extracted color and shape features, respectively. The combination of color and shape features could increase the accuracy rate of classification, especially for potato tubers and clods. The overall accuracy rate was 97.8% in 2016 and 98.1% in 2017. It was evident that the color features dominate in the classification model. Shape features based on the correcting image showed positive effect in classification. It turned out that the combination of shape and color features can obviously improve classification performance.