The availability of new sensor technologies, such as thermal and hyperspectral imaging, enables early-stage weed detection and species identification and density estimation, both of which are crucial for effective weed management. Thermal imaging successfully distinguished between dicotyledonous (oilseed rape, pea, Stellaria media, Triplerospermum inodorum, Veronica persica) and monocotyledonous species (barley, wheat, sorghum and Echinochloa crus-galli) except Amaranthus retroflexus, during early growth stages. The most pronounced differences in hyperspectral reflectance occurred at 550 nm, where five distinct plant groups were recognisable (sum of squares = 0.7604, F-value = 105.1). The highest hyperspectral reflectance was recorded for oilseed rape, followed by Stellaria media. The same trend was found for the normalised difference index (NDI), which also showed five distinct groups. These findings indicate that thermography and hyperspectral imaging have strong potential as effective tools for supporting weed detection in precision agriculture; however, further research and field validation are required before routine implementation in agricultural practice.
Vašková et al. (Tue,) studied this question.