In this paper a simple algorithm for flight targets recognition with combination of radiant intensity and gradient in multispectral image is presented by using the continuous characteristics of radiant intensity of adjacent infrared band for man-made flight targets.This algorithm filters the multispectral image with a highpass filter to suppress the background.The probability density functions(PDFs) of intensity threshold and gradient threshold of spectrum radiation are constructed based on the assumption that the retained image follows the Gaussian distribution.And based on these PDFs,the 3σ rule is obtained to detect and identify the target and bait.The dual thresholds based on the statistical characteristic of data enhance the adaptive capability of the proposed algorithm.The simulation on targets recognition with the proposed algorithm example in strong noise is conducted,and results show the effectiveness of the proposed algorithm.