Detection and quantification of broadleaf weeds in turfgrass using close-range multispectral imagery with pixel- and object-based classification

The current practice used to evaluate broadleaf weed cover in turfgrass is visual assessment, which time consuming and often leads inconsistencies among evaluators. In this study, we investigated the effectiveness of constructing Random Forest models (RF), either pixel-, object-based (OBIA) or a combination both detect quantify cover. High resolution multispectral images were captured 136 plots, seeded with five species Festuca L. overseeded clover (Trifolium repens L.), daisy (Bellis perennis yarrow (Achillea millefolium mixture all three weeds. Ground measurements vegetation bare soil taken point quadrat digital image analysis. Weeds detected 99% accuracy by OBIA, followed combined approach (98%) Pixel-based (93%). Accuracy at distinguishing was somewhat lower (89%, 81% 90%, respectively), contributing most decrease accuracy. predictions based on ground further compared field measurements. For classification, that shape features (OBIA combined) resulted better agreement Pixel- classifications. Our study suggests comprised such as can be accurately quantified high images; however, quantifying remains challenging.