Dipartimento di Ingegneria informatica, automatica e gestionale

Timely weed mapping in crop post-emergence situations is a challenging task required for developing precision weed management solutions. It is necessary to discriminate the crop from the weeds and, if possible, to distinguish different weed species. The ability to map weeds using hyperspectral images acquired from an unmanned airborne vehicle (UAV) over a maize field was evaluated by comparing different classification strategies. The results were mainly affected by the variability in crop and weed spectral signatures. The discrimination between maize and weeds allowed the quantification of their relative ground cover, showing moderate relationship with their relative leaf area index.