Development of a smart unmanned system hyperspectral imager for ground fuel characterization

Abstract

Hyperspectral imaging and machine learning have been shown to be valuable tools for environmental monitoring but present a variety of issues. Hyperspectral imaging generates a significant amount of data that needs to be stored for gathering results, and most environmental monitoring occurs offline and in remote locations. This presents a need for some way to classify hyperspectral images in real time without the need to store the large data. Our solution is to create a custom imager that contains the necessary hardware for collecting hyperspectral images and implementing the image processing and classification onto the same system. This allows for the classification of hyperspectral images without the need to store any data for processing later. The development of both custom hardware for interfacing a complimentary metal-oxide-semiconductor (CMOS) sensor with a system-on- module field-programmable gate array and various peripherals alongside custom firmware and embedded software will allow for a singular embedded system capable of real-time classification of hyperspectral data in the field.