Drought Injury Detection in Turfgrass

Classification map of turfgrass plot. Red indicates false negative for the actual ET replacement level. Green indicates true positive.

Research goal

  • Improve drought injury detection in turfgrass using aerial multispectral imagery
  • Can multispectral cameras detect drought injury that can not be seen in the visible spectrum?
  • Can we detect drought injured turfgrass earlier before the effects are seen in the visible spectrum?

Personal involvement

  • Develop pre-processing and machine learning pipeline for drought injury classification in turfgrass with multispectral images using Python and XGBoost
  • Compare multispectral classification performance with RGB images to understand detection performance gain using all spectral bands

For a percent green calculator, I have written a script to process directories of lightbox images in Python for those who may not use ImageJ/Fiji regularly.

Kyle Cheung
Graduate Student

Working on applying advanced remote sensing and AI to next generation of agricultural decisions.