Abstract: To improve the accuracy of coal fire detection in a mine, RGB and thermal infrared images were respectively collected at daytime and night time using UAV equipped with color and thermal infrared camera. Based on object oriented classification method, color orthophoto images were classified in mining area and the emission rate of corresponding categories were given. The thermal infrared image is inlaid as a positive projective image after radiation calibration, the land surface temperature is inversed according to the radiation conduction equation and the Plank function, and the thermal anomaly extraction algorithm is used to identify the coal fire zone. The test shows that the overlap ratio of the coal fire spot and the thermal infrared detection by unmanned aerial vehicle is 96.72%, which indicates that the accuracy of thermal infrared remote sensing coal fire detection method by unmanned aerial vehicle is reliable and the technology is feasible.