AI-Backed Off-the-Shelf Drones Give Farmers a Heads-Up on Optimal Harvest Times
Automated data collection allows for growth projections, researchers find, automatically pinpointing the best harvest time.
Researchers from the University of Tokyo and Chiba University have come up with a way for farmers to boost vegetable yields — by deploying drones and artificial intelligence to analyze individual crops' growth characteristics.
"The idea is relatively simple, but the design, implementation and execution is extraordinarily complex," says Wei Guo, associate professor at the University of Tokyo's Laboratory of Field Phenomics. "If farmers know the ideal time to harvest crop fields, they can reduce waste, which is good for them, for consumers and the environment. But optimum harvest times are not an easy thing to predict and ideally require detailed knowledge of each plant; such data would be cost and time prohibitive if people were employed to collect it. This is where the drones come in.”
The team's "relatively simple" idea is this: automating data collection for large-scale farming by using low-cost drones equipped with cameras feeding into a machine-learning system which can analyze the growing plants and make predictions regarding their growth characteristics. While the study involved broccoli, the team believes the approach is generalizable to a range of crops.
"It might surprise some to know that by harvesting a field as little as a day before or after the optimal time could reduce the potential income of that field for the farmer by 3.7 percent to as much as 20.4 percent," Guo claims. "But with our system, drones identify and catalog every plant in the field, and their imaging data feeds a model that uses deep learning to produce easy-to-understand visual data for farmers. Given the current relative low costs of drones and computers, a commercial version of this system should be within reach to many farmers."
The team's field tests used off-the-shelf DJI Phantom 4 v2, Phantom 4 RTK, and Mavic 2 Pro drones, capturing 5,472×3,658-resolution 2D images and feeding them back to a desktop computer with a pair of NVIDIA GeForce GTX 1080 Ti graphics cards. These churned through the data using the YOLO detection and BiSeNet segmentation models, resulting in a projection for optimal harvest time.
The team's work has been published in the journal Plant Phenomics under open-access terms.
Freelance journalist, technical author, hacker, tinkerer, erstwhile sysadmin. For hire: freelance@halfacree.co.uk.