3D Gaussian Splatting for Agricultural Phenotyping: A Feasibility Study on Crop Morphology Reconstruction

High-throughput crop phenotyping plays a vital role in precision agriculture, particularly in addressing critical challenges such as yield improvement. However, existing methods face a persistent dilemma between accuracy and cost. While multi-view RGB-D systems like Pheno4D achieve sub-millimeter precision, their reliance on expensive camera arrays (approximately $100,000) and offline processing severely limits field deployment. The 3D Gaussian Splatting (3D-GS) technique, known for its real-time radiance field rendering capabilities, has been widely adopted in medical imaging and urban scene reconstruction. Nevertheless, its application to agricultural phenotyping remains unexplored. This study systematically validates the potential of 3D-GS for crop phenotyping through targeted experiments.