Bidirectional reflectance distribution function (BRDF) is extremely critical for quantifying anisotropic reflectance properties of land surfaces. However, traditional satellite-based BRDF products often lack the spatial resolution and timeliness required for precise large-area modeling. Unmanned aerial vehicle (UAV) technology offers a solution by enabling efficient multi-angle data acquisition at high resolution.

In a study published in ISPRS Journal of Photogrammetry and Remote Sensing, a research group led by Prof. LI Haiwei from Xi’an Institute of Optics and Precision Mechanics of the Chinese Academy of Sciences proposed a new approach, the adaptive generalization-driven methodology, which aims to obtain high-precision BRDF modeling using UAV-borne multispectral remote sensing.

Researchers employed a DJI M600 drone equipped with a Red-Edge multispectral imager to implement a multi-rectangle nested flight scheme. This drone captured pixel-level multi-angle data across five spectral bands for six feature types. To ensure the data accuracy, illumination variations were monitored in real-time using an irradiance measurement device, while slope and aspect were derived from reconstructed maps.

Based on the comprehensive dataset, researchers established a standardized multi-angle information database integrating imagery, terrain, and lighting conditions, which is expected to create a robust foundation for precise modeling. Therefore, the three widely used surface reflectance models could be enhanced to better accommodate complex variables like illumination dynamics and terrain undulations. 

Experimental results showed that the corrected BRDF models reduced fitting errors significantly, with root mean square errors (RMSE) as low as 0.0109. The adaptive diffusion method identified the optimal observation areas for each feature. This study enables centimeter-to-kilometer scale reflectance characterization across diverse landscapes.

"The integration of UAV data with adaptive BRDF modeling opens up new possibilities for high-resolution, large-scale remote sensing applications. And this method not only improves classification accuracy but also enhances the efficiency of reflectance inversion in complex environments,” said Prof. LI.