Lithological classification based on Gabor texture image analysis

Abstract

Lithological classification is important to improve control of the grinding process in a mining plant. Based on the lithological classification the hardness of the mineral can be estimated and mill operation can be optimized. In this paper we proposed a method for rock lithological classification based on Gabor texture analysis and support vector machine classification. We use images from a database formed using rocks extracted from a typical mining plant in Chile. Six different lithological classes were used. Ten images for each lithological class were used for each class for training and another ten images for each class was used for testing. Results on the testing database were measured using 5 cross-validations for the six classes. Our results show that extracting texture features with Gabor filters and rock segmentation based on the Watershed transform reached over 80% accuracy on the testing database.