Utility of Hyperspectral Reflectance for Differentiating Soybean (Glycine Max) and Six Weed Species

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

Abstract

Various analysis methods were evaluated for their utility in using hyperspectral reflectance data to discriminate between 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. Vegetation indices were created and combined with principal component analysis and linear discriminant analysis to determine their classification accuracy. Principal component analysis using vegetation indices produced the poorest classification accuracies, generally less than 50%. Best spectral band combination (BSBC) provided the greatest classification accuracies, with accuracies greater than 80%. Additionally, the BSBC suggested three areas of interest for species discrimination. These areas were located at 1445 to 1475 nm, 2030 to 2090 nm, and 2115 to 2135 nm. The top ten spectral bands determined by BSBC were combined with vegetation indices and re-analyzed using principal component analysis and linear discriminant analysis. This increased classification accuracies above and beyond those obtained by use of vegetation indices alone, suggesting greater crop and weed species differentiation can be obtained when using sensors that include these regions of the electromagnetic spectrum.