A Validation Case of a Diffuse Light Correction Form of the Kernel-Driven Model Using UAV Multiangle Observations

Accurate target characterization extraction under overcast conditions is a great challenge for all-weather quantitative remote sensing. Although existing atmospheric correction methods based on optical scattering models can effectively mitigate image interference, their performance often depends on numerical simulations or idealized atmospheric assumptions, lacking systematic validation in real and complex environments. In this study, we assess the performance of a diffuse light correction (DLC) form of the Ross-Li kernel-driven bidirectional reflectance distribution function (BRDF) model using multiangular and multispectral reflectance data acquired by the DJI P4M platform over identical surfaces under both overcast and clear skies. Experimental results demonstrate that the model significantly suppresses the smoothing effect of anisotropic reflectance features near the hotspot, and the BRDF inversion accuracy is comparable to that obtained under clear-sky conditions with root-mean-square errors (RMSEs) less than 0.02 on average. This study validates the practicality and effectiveness of the improved kernel-driven model using field observations for the first time, providing critical data and theoretical insight for advancing all-weather operational remote sensing applications.