The Utility of Gordon’s Standard NIR Empirical Atmospheric Correction Algorithm for Unmanned Aerial Vehicle Imagery

Unmanned aerial vehicle (UAV) imaging has been increasingly applied for environmental monitoring due to its high spatial resolution. However, the digital number representations of UAV images cannot precisely represent the true radiance of ground objects due to the complexity of solar radiation transmission through the atmosphere. Nonetheless, previous studies have not considered the radiometric information of UAV images in terms of the atmospheric effects obtained under different meteorological conditions. This is addressed in the present work by proposing an atmospheric correction algorithm for UAV imagery based on Gordon’s standard near-infrared (NIR) atmospheric correction applied to water color remote sensing. First, the atmospheric path radiance ( L path ) and the diffuse transmittance ( t ) of water bodies in each spectral band of UAV images are obtained by measurements of the radiance of clear water bodies. Then, the obtained values of L path and t are employed in the atmospheric correction processing of UAV imagery. As such, the proposed method is generally applicable to areas with clear water bodies. The atmospheric correction performance of the proposed method is compared with that of dark pixel atmospheric correction. The vegetation radiance corrected by the proposed correction ( L v ' ( λ i )) was more closer to the measurement than that corrected by dark pixel correction ( L v '' ( λ i )). Three vegetation indices (ExG, NGRDI, and NDVI) were used to further compare the performance of the two atmospheric correction. The average percentile differences of EXG (12.41%, 19.15% and 10.18%) obtained from the UAV images corrected using the proposed algorithm ( AC_W ) were less than those (48.18%, 18.75%, and 58.9%) obtained from the UAV images corrected using the dark pixel method ( AC_D ). So did those of NGRDI and NDVI. The proposed method is demonstrated to perform distinctly better. Despite some error (5–7%), the proposed method provides an alternative for applying an atmospheric correction to UAV imagery under different meteorological conditions. The proposed method is also demonstrated to be simpler and more operable than the atmospheric aerosol optical thickness method.