Abstract: In order to improve the hierarchical management and control of coal mine safety risks and the ability of accident prediction and early warning, and realize the dynamic supervision and risk prevention and control of enterprises in the coal mine industry, we have established the development platform architecture and business application architecture of the coal mine intelligent video early warning system by using video analysis technologies such as artificial intelligence and machine learning to intelligently monitor many working scenes on and under the shaft, automatically detect hidden dangers and alarm, Put forward the index system of intelligent video early warning analysis of coal mines, early warning prevention and control and analysis methods, and establish a unified information platform for safety production; Based on video image data, seven application scenarios of intelligent video early warning system are analyzed, including GIS map dynamic panoramic display of working area, standardized wearing of safety helmets, identification of railway station personnel, identification of crawling and pedaling transport vehicle behavior, identification of key post personnel off post, belt deviation detection, belt empty load detection, and underground fireworks detection.