Remote sensing weeds, cover crop residue, and tillage practices in soybean

Research was conducted to evaluate the potential of remote sensing for detecting weeds in soybean. The first objective was to evaluate the potential of multispectral imagery for detection of late-season weed-infestations. The second objective was to identify bands derived from hyperspectral data capable of discriminating small, early-season pitted morningglory intermixed with soybean from weed-free soybean in different tillage and cover crop residue systems. The final objective was to determine the success of wavelet-based analysis of ground-collected hyperspectral data for discriminating reflectance patterns of pitted morningglory intermixed with soybean from weed-free soybean.; Late-season weed infestations were discriminated from weed-free soybean with at least 90% accuracy when using green, red, and near-infrared spectral bands and a normalized difference vegetation index as classification variables in discriminant analysis. Discriminant analysis functions formed for one data set and tested on another set were 81 to 90% accurate in discriminating weed-infestations when soybean growth stage between data sets did not differ substantially.; For the second objective, reflectance data were collected between 350 and 2,500 nm in 2151 discrete bands, with spectral bandwidths of 1.4 nm between 350 and 1050 nm and 1.0 nm from 1000 to 2500 nm. Eight 50-nm bands (1 ultraviolet, 2 visible, 4 near-infrared, 1 mid-infrared) were used to classify reflectance properties of pitted morningglory intermixed with soybean from weed-free soybean with 83 to 100% accuracy within the different tillage/residue systems. Pitted morningglory growth stage influenced detection capabilities more than tillage and residue systems. The ability to discriminate pitted morningglory intermixed with soybean from weed-free soybean was attributed to differences in reflectance properties of pitted morningglory and soybean, not from increase in total vegetation biomass in the pitted morningglory intermixed with soybean plots.; Thirty-six mother wavelets were compared for their ability to identify wavelet detail coefficients that could be used to discriminate reflectance properties of pitted morningglory intermixed with soybean from weed-free soybean (2 class system). Reflectance data were collected in the same manner as in the second objective. The two classes were discriminated with at least 87% accuracy, regardless of mother wavelet, when pitted morningglory was in the 2- to 4-leaf growth stage. Most wavelet coefficients were derived from red and near-infrared portions of the spectrum.