Low-Cost Dispersive Hyperspectral Sampling Scanner for Agricultural Imaging Spectrometry

A hyperspectral (HS) imaging system in the 400-1650 nm range was developed based on a point spectrometer matched to a double-wedge prism scanner (Risley prism). This type of scanner is efficient and robust in monochromatic laser systems operating under field conditions. Here, we adopted this scanning principle to develop an inexpensive replacement for the HS camera, suitable for application in precision agriculture and environmental monitoring. However, the application of such a scanner in HS imaging is challenging, due to the broadband nature of HS imaging and the dispersive nature of prisms. The sampling characteristics exhibit wavelength-dependent and space-variant behavior, which affect both ground pixel size and location. In this work, we modeled light propagation through the proposed system using a ray-tracing approximation. Based on this model, we proposed a three-step calibration process. A prototype was built and optimized, followed by an intensive validation process that showed a high correlation between the scanner's performance and model prediction of ground sample diameter, as well as accuracy in the spectral measurement of the system. Testing reflectance measurement accuracy showed a normalized sum of absolute difference (NSAD) of 4.24% and less. The calibration process was demonstrated in simulations. Finally, preliminary field experiments with two spectrometers in the range of 400-899 nm (Vis-NIR), and 950-1650 nm shortwave infrared (SWIR) showed that the proposed system can support outdoor monitoring, thus having the potential capabilities for supporting agricultural monitoring tasks.