Abstract: The flammable liquid detection, especially the emote flammable liquid dynamic monitoring, is of great importance to public security, customs and civil aviation. Hence, a near infrared (NIR) multi-spectral imaging system based on liquid crystal tunable filter was developed. To increase the speed of scanning imaging and data processing, the number of spectral channel was reduced remarkably by band selection method. The experiment of flammable liquid detection by the system was carried out in different environments, and the result show that the detection accuracy of the system is 100% when the object was motionless, and it is more than 95% when the movement speed of object was less than 0.2 m/s.