Chris Reberg-Horton is confident robots will eventually come to the farm, but said simple camera technology can immediately change farming.
At a Glance
- Computer vision to revolutionize field monitoring
- Experts crucial in training AI for agricultural tasks
- Swarms of small robots to replace large machinery
Beyond doubt, artificial intelligence is now the driving force in precision agriculture. But regarding its impact, two questions come to mind?:
-
When will the technology arrive?
-
How dramatically will farming change?
The topic is top of mind for Chris Reberg-Horton, professor of cropping systems at North Carolina State University. He is also co-director of Precision Sustainable Agriculture, a network of on-farm and on-station experiments that utilize the latest developments in sensors, IoT platforms, and machine learning to collect and standardize field data from a wide range of agricultural environments.
The network operates in 25 states with more than 120 locations.
‘Camera-everywhere’ concept
While Reberg-Horton is confident robots are coming to the farm, he stressed that they won’t be planting, harvesting, and controlling pests for another 15 to 20 years. He noted that computer vision, or simple camera technology, is here now and can immediately change farming.
“It’s going to be that ‘camera everywhere’ concept. I think there should be cameras strapped to every piece of machinery you bring into a field. They are so low cost now that there is no reason that you should not put them everywhere,” Horton told peanut scientists at the 57th annual meeting of APRES (American Peanut Research and Education Society) July 16 at the Omni Hotel in Richmond.
“They are not meant for specific tasks. They are meant for all tasks: diseases, fertility, insects, weed infestations, and should be trainable for these camera systems. Every time a farmer is moving through a field, we are mapping that field for stresses we can detect,” Reberg-Horton said.
Teaching the machine
Reberg-Horton told the conference of peanut researchers and educators that they will play a key role in helping farmers adopt computer vision and robotics into their operations. He said farmers will soon be able to target the technology to specific tasks, crops, and pest challenges. He said agriculture is “100,000 different cases that have to be done.”
“Computer vision isn’t just about counting things. It can do a whole lot more things. The key to computer vision is experts annotating images. We have to teach the machine. You’re not coding it; you’re teaching it like you teach your dog to do something,” Reberg-Horton said.
Technology, though already precise and sophisticated, is changing rapidly. Reberg-Horton said artificial intelligence will need a minimum of 5,000 images to perform a task.
For example, Reberg-Horton said computer vision is so good now that it cannot only detect the difference between clover and grass in a field, but it can detect the difference between red clover and white clover. However, thousands of images annotated at a very precise level are needed for the system to work.
Equipment changes
Reberg-Horton said drones and robots coming to the farm in the future will allow farmers to rely on much smaller equipment rather than huge tractors, combines, and spray rigs. He said equipment became so big in the first place because farmers were trying to reduce the number of operators driving the equipment.
“It is really expensive to put an operator out there. But once we’re autonomous, things will be moving through the field on their own. What seems like a better investment? One 120-foot sprayer, or maybe six of them that are only 20 feet wide or maybe 10 that are 12 feet wide? The question of how small a robot should be is going to be a fundamentally different question and a different answer than how big equipment is currently,” Reberg-Horton said.
“We are going to shrink down to the smallest size that does the job. Big farms will get a bunch of them. Small farms will get fewer of them. There will be swarms of them, and you will have an economy of scale,” he said.“John Deere won’t have to make 12 different sizes of these things. They will make one. The bigger you are, the bigger the swarm your order will be. It will work night and day. It will cover a whole lot of ground. If you lose one or two, whatever, there is a ton of them out there, whereas if your million dollar combine goes down, you got a problem.”
About the Author
