The Semantic Segmentation Method of Coal Rock Image Based on ResNeSt

Chinese coal industry is n important basic industry supporting national economic construction and ensuring social development. It always provides a solid guarantee for the country's energy security. Coal rock identification is one of the key technologies to promote the intelligent construction of the coal industry. At present, compared with image recognition technology, image semantic segmentation technology can more clearly distinguish coal and rock parts in coal and rock images. It is helpful for the identification work of fully mechanized mining faces. This paper takes the semantic segmentation of coal and rock images as the main research goal. A coal-rock image semantic segmentation model ResNeSt-Unet is designed. It solves the problem that a large number of pooling operations in the U-net network affect the ability to extract image features. The ResNeSt-Unet network uses a ResNeSt module with a joint attention mechanism and a sampling module with a channel attention mechanism based on the U-net network. It improves the network's ability to extract coal and rock images. Finally, two different evaluation indexes and standards are used to compare and analyze the experimental results. Experiments show that the PA and IOU of the ResNeSt-Unet network are 96.62% and 94.83%, respectively. Compared with other networks, it has certain advantages.