Huajian Liu, Bettina Berger and Reddy Pullanagari

Australian Plant Phenomics Facility, The Plant Accelerator®

School of Agriculture, Food and Wine

University of Adelaide

Hyperspectral imaging offers numerous advantages in plant phenotyping and precision agriculture when compared to standard RGB (red-green-blue) imaging. However, the utilization of hyperspectral imaging in both remote and close-range sensing necessitates consideration of diverse factors, an aspect that has not received adequate attention within the research domain.

In the context of long-range scenarios, terrestrial vegetation can be reasonably approximated as planar entities. Key factors influencing imaging outcomes encompass the geometry of sun-cloud-sensor, atmospheric conditions, and soil backgrounds. Conversely, in close-range contexts, plants cannot be simplistically regarded as planar objects. Variables, such as imaging distance, surface angles, shadow interplay, specularity, as well as neighbouring leaf-driven aspects of reflectance and transmittance wield substantial impact on imaging results.

Furthermore, illustrative instances are furnished to underscore disparities between hyperspectral imaging executed at remote distances versus close range. Ultimately, the discourse delves into the research trajectory concerning data calibration, optimization, and the integration of deep learning techniques.