Development of image-based data processing methods for UAVs in peanut and cotton production

Authors: YOUNG, WADE - Oklahoma State University; LONG, JOHN - Oklahoma State University; Bennett, Rebecca; WANG, NING - Oklahoma State University; WECKLER, PAUL - Oklahoma State University

Submitted to: American Society of Agricultural and Biological Engineers
Publication Type: Abstract Only
Publication Acceptance Date: 3/29/2021
Publication Date: N/A
Citation: N/A

Interpretive Summary: Unmanned Aerial Vehicles (UAVs) have become a major asset to agricultural producers and consultants all over the world. However, much work remains in applying UAV technology to agricultural research. A custom-built UAV was developed to aid various peanut and cotton research projects at Oklahoma State University’s Caddo Research Station in 2020. Using readily available and affordable cameras, thermal, near-infrared, NDVI, and RGB images of the crops were collected during critical periods of the research projects. In addition to the custom-built UAV, a commercially available drone equipped with a high-resolution RGB camera was used to collect data. Multiple methods for analyzing and visualizing the image data were developed. The processed cotton data were used to predict the percentage of open bolls and visible cotton, with the goal optimizing timing for harvest-aid chemical applications. The processed peanut data were used to locate diseased plants and predict the severity of disease and compared to ground truth data collected by a pathologist. In 2021, field tests will be conducted to verify the performance of the developed data analysis algorithms.

Technical Abstract: Unmanned Aerial Vehicles (UAV’s) have become a major asset to producers and consultants all over the world. A custom-built UAV was developed to aid various research. In 2020, both peanut and cotton crops were monitored with the developed UAV system at Oklahoma State University Caddo Research Station in Southwestern Oklahoma. Using readily available and affordable cameras, crop data were collected during the crop growth periods which included thermal, near-Infrared, NDVI, and RGB images. In addition to the custom-built UAV, a DJI Phantom 3 Professional equipped with a high resolution RGB camera was used to collect data as well as replications. Multiple methods for analyzing and visualizing the image data were developed using Python with the OpenCV and Numpy Libraries. The processed data on cotton were used to predict the percentage of open boll and percentage of visible cotton to aid in the determination of timing for harvest-aid chemical applications. The processed peanut data were used to locate various diseased plants and predict the severity of disease damages based on NDVI, thermal, and RGB images for the crop across the entire growing season. The results were compared with ground truth data collected by peanut specialists. In 2021, field tests will be conducted to verify the performance of the developed data analysis algorithms.