Abstract: In view of the fact that few existing coal-rock recognition methods achieve a satisfactory performance when training samples are insufficient,an effective coal-rock recognition approach based on completed local binary patterns (CLBP) and support vector guided dictionary learning was proposed. It includes four steps as follows: Firstly, the multi-scale CLBP-based feature vectors of the coal and rock images were obtained. Secondly,support vector guided dictionary learning was conducted on the feature vectors of training samples. After that,one dictionary for coal-rock characterization,several weight vectors and biases for coal-rock identification were earned. Thirdly,the representation of the multi-scale CLBP feature vector of the test sample with respect to the dictionary (i. e. ,the code vector) was ac- quired. Finally,the categorization regarding the code vector was completed by using discriminant function. Experimen- tal results demonstrate that compared with other existing approaches,the proposed one yields a higher correct recogni- tion rate. Furthermore,it has a very high recognition accuracy even in sample-insufficient random sampling experi- ments. The time-consuming dictionary learning involved in the proposed approach scarcely hampers its realtimeness. The storage requirement for this approach does not depend on the number of training samples,which facilitates its hardware implementation in the future to some extent.