Early season weed mapping in sunflower using UAV technology: variability of herbicide treatment maps against weed thresholds

摘要: Site-specific weed management is defined as the application of customised control treatments only where weeds are located within crop-field by using adequate herbicide according to emergence. The aim study was generate georeferenced seedling infestation maps in two sunflower fields analysing overlapping aerial images visible and near-infrared spectrum (using or multi-spectral cameras) collected an unmanned vehicle (UAV) flying at 30 60 m altitudes. main tasks focused on configuration evaluation UAV its sensors for image acquisition ortho-mosaicking, well development automatic robust analysis procedure mapping used design a site-specific program. strategy based seven thresholds with 2.5 steps increasing ratio from 0 % (herbicide must be applied just when there presence absence weed) 15 % coverage >15 %). As first step imagery analysis, rows were correctly matched ortho-mosaicked imagery, which allowed accurate object-based [object-based-image-analysis (OBIA) methods]. OBIA algorithm developed successfully classified sunflower-rows 100 % accuracy both all flight altitudes camera types, indicating computational analytical robustness OBIA. Regarding discrimination, high accuracies observed any altitude, highest (approximately 100 %) being those recorded threshold, although satisfactory results 5 % also observed, higher than 85 % field 1 2. lowest (ranging 50 60 %) achieved threshold. Herbicide savings relevant fields, they 2 due less infestation. These varied different scenarios studied. For example, 30 m altitude camera, range 23–3 % (i.e., 77 97 % area) could treated 0–15 % thresholds. computed multiple data permitted calculation requirements timely post-emergence management.