Robert Green, Ph.D., of Cal Poly Pomona's Plant Science department recently embarked on a collaboration with the departments of computer science and aerospace engineering to see how unmanned aerial vehicles (UAVs, commonly called drones) can be used to assess turfgrass health and quality. Plots of bermudagrass at a golf course in Riverside, Calif., and at the university's Center for Turf Irrigation and Landscape Technology were utilized.

Eight models were programmed to learn how to assign a number rating for turfgrass quality and color on a scale from 1 to 9 (1 being dead and brown, 5 being minimally acceptable, and 9 being maximum quality or color). Training for these machine learning models took anywhere from a couple of hours to a couple of days.

Using digital image analysis, 8,301 sub-images captured by a camera were examined; some of the images were taken manually and some were taken by a drone.

"The model had an error of around 1.0," said Green, whose error rate is in the 0.3-0.4 range. "It's a good start, but there's more work to be done."

"It can be a real time automatic download, with somebody sitting at a computer that can see what the visual rating is too – after the model has been adequately trained," he continued. "That would be very handy."

"Most people working with turf and drones and calculating normalized difference vegetation index," a ratio that indicates turfgrass stress and digital image analysis for green color, explained Green. "This concept of using deep learning with the drone images, I haven't seen anybody else doing that on turf."

Since this study has completed results, similar studies are now being conducted to observe different plants.

"Drones are fast and can do more than humans on their own. When you have hundreds of acres of something, it's hard to observe all that – not to mention the economics of it, too," Green said. "The big picture is that drones can be used for so much more. The trick is understanding what the information means in terms of landscape and plant maintenance and production."