We are appropriately accustomed to thinking that image sensing is one of the most sophisticated techniques to investigate the environment due to the large amount of information carried by the two(three)-dimensional signals. However, image technology needs calibration, setup, and environmental embedding requirements that are frequently critical. On the other hand, spectral sensing still allows a massive amount of information (each point of the spectrum) using a single transducer.

Recently, investigations on novel bands (mm-wave, terahertz radiation) together with traditional ones (ultrasound, optical, microwave) opened new perspectives in Precision Agriculture (PA), permitting complex and deep physical analysis of materials, an aspect where image sensing techniques lag behind.

The main problem of spectra sensing is understanding where the information is hidden in spectra and how to extract it in a measurable form. For this reason, in this presentation, we will introduce Multivariate Analysis techniques showing the power of this approach in low-power, low-computation edge computing applications and perspectives in PA.