Cotton Boll Counting with LIDAR

Thursday, January 4, 2018: 4:00 PM
Salon D (Marriott Rivercenter Hotel)
Andrew N. French , USDA-ARS
Alison L. Thompson , USDA-ARS
Michael A Gore , Cornell University
Terrestrial Light Detection and Ranging (LIDAR) has the potential to provide accurate boll counts and maps of boll distribution over a defoliated cotton canopy. Using a field-portable LIDAR unit and GPS receiver, cotton plants can be scanned plot-by-plot, returning point cloud data that can be aggregated and mapped. Since 2013 small plot LIDAR experiments have been conducted at Maricopa, AZ to develop acquisition and data processing systems. Results, achieved using multiple mobile platforms- high clearance tractors, hand-push phenotyping carts, and semi-robotic rail carts- have shown that cotton bolls can be mapped and counted provided, platform travel speeds are slow (~0.5 m/s), GPS locations are accurately synchronized with the LIDAR, and cotton plants are fully defoliated. However, obtaining consistent results has been difficult because LIDAR scan rates were too slow, leading to spatial sampling gaps and inaccurate sensor positioning. Experiments in 2017 were undertaken to reduce these problems by deploying a LIDAR unit capable of scanning at rates exceeding 200 Hz. The different model, mounted on a motorized cart, greatly improved spatial sampling compared with prior experiments. Using data collected October-November 2017 over a ~1 ha RBTN trial, boll counts and boll distribution results will be presented. An assessment of hardware and software accomplishments and remaining challenges will be discussed.