Abstract: A coal-rock image feature extraction and recognition method based on binary cross-diagonal texture matrix was proposed. Binary cross-diagonal texture matrix of coal-rock image is extracted firstly. Then feature vector of coal-rock image is constructed by angular second moment energy, relevance, variance, inverse difference moment, entropy, sum entropy, difference entropy, sum average, contrast, inertia moment and information measurement of correlation, which are extracted from the binary cross-diagonal texture matrix. Finally, sparse representation is adopted to recognize coal-rock images. The experimental results show that the method can achieve better performance than image feature extraction and recognition method based on cross-diagonal texture matrix, whose average recognition rate can reach 94.38%, and improve real-time performance of coal-rock recognition with shorter feature extraction time of single image.