Reflected Intensity Correction Based on Global 3-D Plant Structure Acquired by Hyperspectral LiDAR

Abstract

The precise quantification of 3-D plant biochemical parameters is essential for ecosystem assessment, yet acquiring accurate target reflectance remains a significant challenge for hyperspectral LiDAR (HSL) due to the strong incidence angle effect induced by complex leaf geometric morphology. Furthermore, existing correction methods are predominantly applied at a single-point scale rather than being strictly limited to it, struggling to accommodate the architectural complexity of whole plants. To address this bottleneck, this study proposes a novel reflectance correction framework for HSL based on the global 3-D structural information of the entire plant to suppress the effects of incidence angle on reflectance. Methodologically, an adaptive neighborhood point determination method based on point density is designed to accurately compute leaf curvature and normal vectors from point clouds, thereby avoiding over-smoothing or geometric distortion. Subsequently, a physically improved Cook–Torrance model for HSL (HSL-CT) is established to decouple diffuse and specular reflectance contributions across 28 spectral bands ranging from 445 to 810 nm. The evaluation results demonstrate that the proposed HSL-CT model consistently outperforms the baseline Lambert–Beckmann (LB) model in overall fitting accuracy and spectral fidelity. By integrating this model with the extracted global structural information, the framework effectively eliminates angular-induced specular over-brightness, significantly reduces spectral errors, and successfully restores native plant spectral signatures. Ultimately, this method can generate incidence-angle-corrected HSL reflectance data with high stability, laying a solid foundation for accurate inversion of biochemical parameters at the individual plant scale.