Architecture and key technologies of coalmine underground vision computing

It has always been a common demand to stay away from the harsh environment with narrow space, numerous devices, complex operation process, and hidden hazards, and realize intelligent unmanned mining in the coal industry. To achieve this goal, it is very necessary for us to develop an effective theory of vision computing for underground coalmine applications. Its main task is to build effective models or frameworks for perceiving, describing, recognizing and understanding the environment of underground coalmine, and let intelligent equipment get 3D environment information in coalmine from images or videos. To effectively develop this theory and make it better for intelligent development of coalmine, this paper first analyzed the similarities and differences about computer vision and visual computing in coalmine, and proposed its composition architecture. And then, this paper introduced in detail the key technologies involved in visual computing in coalmine including visual perception and light field computing, feature extraction and feature description, semantic learning and vision understanding, 3D vision reconstruction, and sense computing integration and edge intelligence, which is followed by typical application cases of visual computing in coalmines. Finally, the development trend and prospect of underground visual computing in coalmine was given. In this section, this paper focused on concluding the key challenges and introducing two valuable applications including coalmine Augmented Reality/Mixed Reality and parallel intelligent mining. With the breakthrough of underground vision computing, it will play a more and more important role in the intelligent development of coal mines.