HS-3D-NeRF: 3D Surface and Hyperspectral Reconstruction From Stationary Hyperspectral Images Using Multi-Channel NeRFs
Researchers' new method captures detailed chemical and shape data of produce without moving cameras.
Deep Dive
Researchers from Iowa State University and collaborators developed HS-3D-NeRF (also called HSI-SC-NeRF), a multi-channel Neural Radiance Fields framework. It reconstructs 3D surfaces and hyperspectral data from stationary cameras by rotating objects in a controlled chamber. The system uses a two-stage training protocol and achieves high spatial accuracy and spectral fidelity across visible and near-infrared bands. This enables automated, high-throughput quality inspection of agricultural produce like fruits and vegetables.
Why It Matters
Enables scalable, automated quality control and phenotyping for sustainable agriculture and food supply chains.