order of magnitude
Although previously reported methods have been mostly applied
to on-ground images, they could also be suitable for the remote
images captured with UAVs, mainly due to the spatial
resolution of on-ground and UAV images being on the same order
of magnitude. Investigations about detailed evaluation of remote
images captured with UAV platforms and their spectral information
or derived vegetation indices with the objective of quantifying VF are currently scarce, although recently Peña et al. (2013) developed
a method for weed mapping in early-season maize fields
using UAV images.
As part of an overall research program to investigate the opportunities
and limitations of UAV imagery in accurately mapping
weeds in early season winter wheat, it is crucial to explore the potential
of generating VF maps from multiple overlapped frames
that were mosaicked as a first step in the proper discrimination
of crop rows and weeds. Such an approach should demonstrate
the ability to accurately discriminate weeds grown between crop
rows to design a field program of ESSWM. Consequently, this work
evaluated the accuracy, spatial and temporal consistency and sensitivity
of different vegetation indices for a wheat crop that were
extracted from visible images acquired with a low-cost camera installed
in an UAV flying. We focused on several acquisition dates
(temporal analysis) and two different flight altitudes. Additionally,
to the best of our knowledge, this is the first work to evaluate the
adequate performance of Otsu’s thresholding method for VF mapping
in UAV imagery.