Abstract: In order to improve the image collection quality of the coal mine robot vision system under harsh working conditions，such as high dust and high humidity underground, a comprehensive dust removal method of self-exposed film and automatic cleaning of the vision lens is proposed, which provides a new idea for the automatic cleaning of the window of the vision system.The authors adopt the dustproof principle of covering the light-transmitting diaphragm on the window surface, realize the automatic peeling of the film and the dust and self-cleaning of the lens window through clever mechanical structure design, fully consider the explosion-proof design requirements of the film peeling device, and develop the virtual prototype design of the film peeling and dust removing system.The system can automatically calculate the cleanliness of the current lens according to the camera exposure and image acquisition quality to determine whether to perform the film removal and dust removal action.At the same time, in the case of film removal and dust removal failure, supplemented by the coal mine robot lens spray dust removal method, the overall architecture of the automatic cleaning and dust removal system is designed.The robot determines the lens spray cleaning strategy through the automatic judgment algorithm of the lens pollution degree.By tracking, automatic positioning and other technologies to realize the docking and cleaning actions of the spray cleaning system and the robot lens.Preliminary experimental research shows that: through online grayscale histogram analysis of the images obtained by the robot vision system, the real-time image degradation degree can be automatically judged, and the dust removal and cleaning device of the vision system can be activated to effectively improve the image quality of the robot vision system.It can provide technical support for coal mine robot underground navigation, positioning and inspection operations.