Abstract: Machine vision system in underground mining has the function of information collection and intelligent decision-making in coal mine intelligenc, but it is affected by coal dust, water mist and light, which leads to unstable image quality and affects the decision-making of intelligent equipment. The research of efficient and reliable cleaning technology is very important to ensure the stable operation of downhole machine vision system. This study introduces the concept, principle and cleaning process of the surface self-cleaning technology of the underground image acquisition system, and analyzes the working conditions and composition of the underground image acquisition system. It is pointed out that the environmental factors of the working face such as dust concentration and illumination conditions have an impact on the path of self-cleaning technology, and the dust production mechanism of coal dust should be clarified to formulate an accurate self-cleaning strategy. The research progress of surface self-cleaning technology in underground mining and conventional conditions is reviewed. Based on the different anti-fouling mechanisms, the research status of three cleaning technologies, jet self-cleaning technology, ultrasonic self-cleaning technology and laser self-cleaning technology, is summarized. The self-cleaning characteristics are analyzed. It is pointed out that high-pressure water jet and surface acoustic wave are suitable for cleaning large-area stains, gas jet, ultrasonic vibration and laser shock are suitable for particle cleaning, and auxiliary multi-functional surfaces will improve cleaning efficiency. Based on the different contact modes, the advantages and disadvantages of direct contact self-cleaning technology and indirect contact self-cleaning technology are compared. It is pointed out that the direct contact self-cleaning technology is simple and efficient, which will damage the lens membrane substrate, and the indirect contact self-cleaning technology has no damage to the base, and the cleaning efficiency is slow. The technical challenges faced by the underground mining machine vision system are pointed out from three aspects: comprehensive self-cleaning strategy, optimization of explosion-proof design and multi-functionalization of lens film materials. Aiming at the scientific problems of multi-technology collaboration and cleaning quality control, a far-near-field collaborative control method is proposed, which can provide relevant technical support for the surface self-cleaning technology of the underground mining machine vision system. Finally, the development trend of surface self-cleaning technology of machine vision system in mining is further prospected. With the application of unmanned mining vehicles to deep ground, deep sea and deep space ( moon ) mining, mining becomes more complicated. It is bound to install more intelligent sensing systems such as machine vision and lidar, and make use of the self-cleaning device of visual system, which can effectively improve the image quality of visual system, so as to ensure the accuracy of unmanned mining.