Abstract: The coal gangue recognition method based on visible light images has low accuracy and slow recognition speed. The coal gangue recognition method based on high-energy ray transmission has significant radiation, resulting in limited application. The infrared thermal imaging has advantages such as strong penetration and no influence from light. But the surface temperature of coal and gangue is relatively close at room temperature, resulting in no significant difference between coal and gangue in infrared thermal images, making it difficult to achieve good recognition results. In order to solve the above problems, a coal gangue recognition method based on water heat transfer and infrared thermal imaging is proposed. The method conducts infrared thermal imaging experiments on coal and gangue under different water temperatures (18, 21, 24, 27, 30 ℃). The method distinguishes between coal and gangue based on the differences in infrared thermal images and temperature changes. The experimental results show that the infrared thermal images of coal and gangue are different at different water temperatures. When the water temperature is lower than the ambient temperature, there is a significant difference between the infrared thermal images of coal and gangue. Under the same water temperature conditions, the difference between the infrared thermal images of coal and gangue gradually increases with time. The surface temperature changes of coal and gangue both show an increasing trend with the increase of water temperature and time. But the surface temperature changes of gangue are greater than those of coal. When the water temperature is 18℃ and the time is 180 s, the difference and temperature difference between the infrared thermal images of coal and gangue reach their maximum. This indicates that low-temperature water can serve as a heat transfer medium, which is more conducive to creating a large temperature difference between coal and gangue. The accurate and rapid recognition of coal and gangue infrared thermal images can be achieved.