关键词: 煤矸石, 图像分类, 反复注意力, 特征提取

Abstract: Accurate classification and identification of coal gangue is an important problem to be solved in the safe and precise mining of coal mines. Residual network shows great advantages in image clas⁃ sification tasks. Using residual network and overcoming its shortcomings in feature extraction，a mine image classification model is proposed. The model combines multi-scale idea and repeated attention method，and also adds ResNet network feature extraction method to the model. In addition，the jump connection is add⁃ ed to the model，which can reduce the calculation amount of the model. Based on the real experimental da⁃ ta，the accuracy of this model is improved by 3 % compared with other classification models.

Key words: coal gangue, image classification, repeated attention, feature extraction