Publication AbstractClose Publication Abstract

Utility of Multispectral Imagery for Soybean (Glycine Max) and Weed Species Differentiation

Prince Czarnecki, J. M., Gray, C. J., & Shaw, D. R. (2006). Utility of Multispectral Imagery for Soybean (Glycine Max) and Weed Species Differentiation. Proceedings 8th International Conference on Precision Agriculture. Minneapolis, MN.

Abstract

Field studies were conducted in 2002 and 2003 to determine the utility of multispectral imagery for identifying soybean, soil, and six weed species which are common problems in production agriculture in Mississippi. Weed species included hemp sesbania, pitted morningglory, palmleaf morningglory, prickly sida, sicklepod, and smallflower morningglory. Planting densities for weeds corresponded to 50, 100, and 200% of published or estimated economic thresholds in soybean. Imagery was analyzed using supervised classification. Bare soil and vegetation were differentiated using a 2-class system. A 3-class system was used to differentiate between bare soil, soybean, and weed. Finally, an 8-class system was used for total discrimination between bare soil, soybean and all weed species independently. Soybean classification accuracies were greater than 95% in the two class system. Greater than 90% accuracies were also observed for pitted and palmleaf morningglory at all but one acquisition date. In the 3-class system, soybean classification accuracies were 70% or greater. Weed classification accuracies increased with density. Pitted and palmleaf morningglory accuracies were greater than 90% at 10 weeks after emergence. In the 8-class system, pitted and palmleaf morningglory again produced the highest classification accuracies at a density of 200% economic threshold, 10 weeks after emergence. All other weed species generally produced less than 50% accuracy, regardless of planting density, however accuracy increased with weed density. These results suggest that multispectral imagery can be of use in situations were discrimination of weed species from each other is not the primary goal, as in the 2- and 3-class systems, but not the 8-class system. The 2-class system could provide assistance in non-selective burndown applications, and the 3-class system could help create site-specific spray maps for a herbicide application in a herbicide-tolerant crop.


Geosystems Research Institute  •  Contact GRI
Modified: December 2, 2020  •  WebMaster  •  Intranet