关键词: 红外热成像技术, 岩石破裂, 张剪裂纹, 目标检测, Faster RCNN

Abstract: Distinguishing rock failure modes is of great significance for mine disaster prevention and control. In order to effectively identify rock failure modes and monitor the failure process,a strain shear crack detection algorithm based on improved Faster RCNN in rock thermal infrared images is proposed. Firstly,the uniaxial compression test of granite was carried out,and the thermal infrared image of rock fracture was obtained by infrared imager,and the infrared fracture data set of rock was made. Then,the Faster RCNN network model is improved,and the feature extraction method is optimized by introducing the attention-guided context feature pyramid network. A cascade structure is adopted to improve the accuracy of detection frame regression. At the same time,the infrared rock fracture data set is used to train and verify the model. Finally,the tensile shear crack identification is carried out in infrared image of measured rock. The experimental results show that the detection accuracy of the algorithm reaches 94. 15%,the detection frame number per second is 19. 27 fps,and the comprehensive performance is significantly improved. The research results can provide a way to predict the rock fracture location and fracture mode.

Key words: infrared thermal imaging technology,rock fracture,tension-shear crack,object detection,Faster RCNN