Recognition and Classification of Coal Sample Composition Using KSVD

In order to solve the problem of coal sample images classification, this paper proposes a new method which includes a pre-processing of coal sample images and an advanced KSVD algorithm. The KSVD algorithm includes sparse coding (find sparse coefficient x)and dictionary update (find dictionary D), then we get an adaptive dictionary iteratively. Finally, we use SVM to solve the problem of coal sample images classification. For the classification of coal samples with no obvious features, the algorithm of this paper has been shown to be effective. Experiments show that the proposed algorithm has good stability and accuracy in classification of coal sample images.