Nonlinear atmospheric correction based on neural network for infrared target radiometry

Infrared radiometry technology is an important means to characterize the infrared signature of targets, and atmospheric correction is a requisite step to obtain the real radiance of targets. A nonlinear atmospheric correction (NLAC) method was proposed to improve the infrared radiometric accuracy for long distance targets in this paper. This method used near-range standard reference source measurement (NRSRM) to calculate the actual atmospheric transmittance and path radiation simultaneously at different locations in a real-time environment. And the theoretical atmospheric transmission and path radiation under the corresponding conditions could be obtained from the atmospheric radiation transmission software as well. Neural network technology was applied to fit the non-linear relationship between them. Thus, the atmospheric transmittance and path radiation over long distances could be predicted to achieve atmospheric correction. Simpler linear atmospheric correction (LAC) and linear enhancement atmospheric correction (LEAC) were also carried out for comparison. The experimental results indicate that the infrared radiometric average error of the proposed method is 6.45%, which is much lower than that of the conventional method, LAC and LEAC that are 16.17%, 11.27% and 7.44%, respectively.