Hyperspectral Imaging for Early Detection of Herbicide-Resistant Weeds in Soybean

April 21, 2021

Prashant Jha, agronomy, Iowa State, and Joseph Shaw, electrical and computer engineering, Montana State, are collaborating on a project that was funded by the Iowa Soybean Research Center in late 2019. The goal of their project, “Hyperspectral Imaging for Early Detection of Herbicide-Resistant Weeds in Soybean,” is to accurately map (drone-based) the location of herbicide-resistant weed biotypes in production fields using advanced optics and computer algorithms.

Greenhouse and laboratory experiments were carried out in 2020 to identify the spectral reflectance of different biotypes of waterhemp plants resistant to ALS inhibitors, atrazine, and/or glyphosate. Hyperspectral imaging and other measurements were taken using artificial light. More plants are being grown from two different species of pigweed (waterhemp vs. Palmer amaranth).

This summer (2021), Jha plans to mount a camera on a drone to collect data in soybean fields with confirmed herbicide-resistant waterhemp populations. This includes imaging herbicide-susceptible and herbicide-resistant weed biotypes at different growth stages to characterize classification accuracies as plants grow. Images will be analyzed to differentiate waterhemp from other weed species in a soybean field and to identify susceptible vs. resistant waterhemp biotypes. A neural network machine learning algorithm will be used to develop classification images for field-scale maps. Using neural networks instead of previously used support vector machine algorithms will improve classification accuracies from 80% to 99%.