A spectral signature-based algorithm for the identifiability of crops and their cultivation conditions

Recent advancements in remote sensing techniques, especially the combination of hyperspectral imaging with analytical algorithms, have greatly improved precision agriculture. This study introduces some algorithms developed for identifying crops and evaluating their growth conditions, focusing on irrigation and fertilisation. The present approach is based on the concept of identifiability of a family of dynamic systems and the differentiation of plants using their spectral signatures. The method uses a repository of spectral data and applies a developed algorithm to compare the measured spectra with the reference database, enabling the identifiability and the recognition of both known and unknown crops. As an application of our approach, we have considered two different crops: mint and rosemary, under different irrigation and fertilisation conditions. The results show that the algorithm achieved a 100% identification rate across the four unknown samples. The minimum spectral distances obtained are 0.01 and 0.03 for rosemary and mint, respectively. Thus, the family of systems was identifiable with a tolerance of η < 0.03. The study concluded that the algorithm effectively classifies the crop type and deduces its growth conditions, demonstrating its effectiveness for agricultural monitoring.