Comparing visible and color-infrared UAV imagery for early-season weed mapping: The case of maize a wide row crop

Abstract

This investigation quantified the efficacy of visible and color-infrared images for weed mapping in a maize field by using remote images captured with an unmanned aerial vehicle at several flight altitudes (from 40 to 100 m). Results on weed discrimination were significantly affected by type of image and by type of weed (broad-leaved or grass weeds). With independence on spatial resolution (i.e., flight altitude), weed coverage was underestimated in the visible images, mostly in the parts of the field infested with grass weeds. On the contrary, analysis done with the color-infrared images was highly precise in the images captured at 40-m flight altitude, although progressively overestimated weed coverage at higher altitude.