New Tech Breakthroughs Are Coming for Cotton
We may not be at Skynet stage yet (Terminator reference check), but there’s no question that the equipment and tech options currently available — as well as those on the horizon — have the potential to make agriculture production more autonomous and efficient.
Based on a conversation with Dr. Ed Barnes, Senior Director, Agricultural & Environmental Research for Cotton Incorporated, here are a few of the tech innovations that just might change the way cotton growers farm in the years ahead.
This is the centerpiece of many of the innovations now hitting or soon coming to market. As Barnes puts it, machine vision can be managed two ways: train the vision system to know what cotton looks like or train it to identify weeds (for weed management applications).
One of those prototypes is coming to market in 2023. John Deere is introducing its See & Spray Ultimate, a factory-installed system available for John Deere 410R, 412R, and 612R Sprayers that enables targeted spraying of non-residual herbicide on weeds among cotton, corn, and soybean plants.
“We’ve been working on this technology since the 2017 acquisition of Blue River Technology,” explains Franklin Peitz, John Deere’s Marketing Manager, Sprayers. “Blue River focuses on machine learning and camera visualization technology along with artificial intelligence to help distinguish between crop and weed. We’ve integrated the sprayer and the technology into a system to work as one. As the machine moves through the field, all of the data is flowing.”
And flowing quickly. The sprayer features a 120-foot carbon fiber boom containing 36 cameras constantly scanning about 2,100 square feet at all times, plus computer processers that identify images of what’s a weed and what’s a crop and signals the correct nozzle to only spray the weed.
“The process of scanning, identification, and spraying is about 0.2 seconds, or the blink of an eye,” says Peitz. “And that’s operating at speeds up to 12 mph in the field.”
The unit – a dual product system featuring two independent tanks – gives growers the flexibility to broadcast one product while simultaneously using See & Spray with a non-residual herbicide out of the appropriate nozzle.
“Some commercial prototypes are already training the system to identify cotton in any situation and spray everything else,” says Barnes. “I think that will be the first thing to come along. Cotton Incorporated is also working on a 5-year project on weed identification with machine vision. If we can train something to find different weeds, it probably works in multiple commodities and also brings the ability to customize what’s applied based on the species of the weed.”
Barnes also noted a company that’s currently offering a machine vision system with camera and processor for corn and soybeans, with preliminary discussions about cotton. As the system passes over the field, it collects images and processes them in real time for scouting disease pressure and insect damage.
With John Deere's new See & Spray Ultimate technology, the process of scanning, identification, and spraying weeds is about 0.2 seconds, operating at speeds up to 12 mph in the field. Photo credit: John Deere
“Related to that and perhaps in a shorter time frame, we’re going to find a way to make better use of UAV imagery,” adds Barnes. “Applying machine vision algorithms with 3D mapping will allow growers to estimate plant height as opposed to just greenness. Studies at Texas A&M have already confirmed proof of concept, and some companies are paying attention to it.”
John Deere’s See and Spray system and SwarmFarm Robotics in Australia are already using machine vision in autonomous weed control systems, primarily for burndown applications at this time. Barnes also mentioned an autonomous system that uses machine vision and lasers to identify and zap weeds (see the video at CarbonRobotics.com). He believes the utility of these and similar systems are going to increase in the future.
There are already several options on the market for autonomous tractors. But what gets Barnes excited is work currently underway at the University of Georgia and Clemson University on more cotton-specific applications.
“Dr. Glen Rains and his graduate students at Georgia have taken open-source software and less than $1,000 in vision and processing equipment and are now able to drive a spider sprayer autonomously that pulls bolls off of a plant,” he shared. “Dr. Joe Maja at Clemson took a commercially available Husky robot (ClearpathRobotics.com) and now has it cultivating down the row. He has made it into an autonomous utility tractor with the smarts to navigate the field and pull smart implements behind it to make it multi-functional.
“To me, all of this means things are ready to happen because the technology is already there.”
Barnes optimistically guesses that 20% of cotton growers are currently relying on electronic record keeping. Programs like MyJohnDeere and FieldView are making it easier to connect equipment for data management — and they are having an impact.
“But I’m also surprised when some really good cotton growers ask me about good software programs to use, because they’re still doing everything in a notebook,” he points out.
Predictive maintenance for equipment is something that’s available now. It provides value to the producer and manufacturer to help avoid equipment failure at critical times of the season.
“I think we’re going to find more ways to make this data more valuable as it becomes possible for the equipment to transmit in real time into the Cloud to make it automatically available,” notes Barnes. “As the data ends up where it’s supposed to, it’s going to be more feasible for growers to use it. They’re going to be able to quantify yields on a regular basis and make a profit map to quickly see the sections of a field that may be losing them money year after year.”
He’s also excited about the potential of increased data options for the ginning industry. TSW Automation has announced plans to offer a turnkey system to make use of current RFID tags to track modules from the field to the gin. Cotton Incorporated is also working with the National Cotton Ginners Association to find ways to integrate the field to gin data with classing data to create fiber quality maps so growers can see how their yields and quality vary from field to field.
Wireless soil moisture monitoring tied to irrigation scheduling tools are now becoming commonplace across the Cotton Belt. Barnes points out that the Smart Irrigation app, developed by the University of Georgia, has proven to be dependable for growers especially in the Southeast and Mid-South. The app plugs into the National Weather Service weather grid and works pretty much anywhere.
“We feel like the crop coefficients are pretty close for the Mid-South and Southeast, and we’ve recently plugged it into the Oklahoma Mesonet,” he says. “We’re still refining it for use in hotter, more arid regions.
For growers who can track variability in yield maps or soil maps, the next step may be adding multiple moisture sensors in the field, which would allow irrigation companies to add wedge control to speed up or slow down pivots automatically.
“There are options,” says Barnes. “But what’s been holding us back is that easy button. Companies are now making integrated systems that tie the moisture sensors to the controls on the pivots. And that’s definitely a plus.”
Cotton Incorporated has been involved in studies in Oklahoma and Georgia on the ability to control downforce at planting. It’s a tool that could provide value by varying the downforce and planting depth when growers face variable moisture conditions at planting and/or changes in soil type.
“The one thing I have trouble getting my head around is how fast some of these planters can go now,” adds Barnes. “I used to think 3-4 mph was fast. Now I don’t think anyone plants under 5 mph — and some are going as fast as 10 mph. The downforce can add value by possibly providing a little more flexibility on the planting window to offset cool, wet conditions that may delay planting.”
