Ecoation’s software learns what a healthy leaf looks like and will detect and mark any leaf that doesn’t meet that criteria. Photo: ecoation
While imaging has been widely adopted in the leafy greens sector, producers growing ornamentals, tomatoes, cucumbers, peppers, and strawberries are also bringing this technology into the greenhouse.
If an experienced grower has one plant in front of them, and their mission is to grow that one plant well, they don’t need imaging to do that. But what about when they have 10 million plants?
Adam Greenberg, CEO of IUNU, explains that imaging and AI is really about helping the grower scale how they already know to grow. The idea is to take established processes and equipment from other areas and adapt them to controlled-environment agriculture (CEA), providing a usable tool to the grower.
“We are bringing widely-used technology from other industries, computer vision, and machine learning, into the greenhouse,” says Mauricio Manotas, Chief Revenue Officer at Ecoation.
What it means is a big step forward for growers. Computer vision and AI are proving themselves as valuable tools in many areas of CEA, particularly for the early detection of pests and diseases.
What Is Computer Imaging?
Ask a tech person about computer vision, and you’ll get an answer using phrases like image segmentation, pattern detection, feature matching, and object detection. Ask an ag-tech person, and they’ll add geo-aware or location-aware to the mix, along with prediction analytics and leaf temperature surface variance. Sounds complicated.
At their core, all systems involve data collection and data analysis. Most are hardware agnostic and operate more like Software as a Service (SaaS) on the cloud. Others are complete package deals. They can be retrofitted to existing greenhouses and vertical farm operations or included in new construction.
“The data analysis or machine learning is the real power behind the system. Massive loads of data without analysis is simply overload,” Manotas explains.
The algorithms sort and rank the data, comparing images to identify areas with issues and suggest the problem: white flies, Botrytis, etc. The data is collected and analyzed into actionable intel. Graphical representations of growth, morning summary emails, and risk maps are provided to the grower. Some systems can forecast future pest and disease levels using current population pressures and environmental conditions. Manotas says these risk maps and forecasts allow growers to get ahead of the issue, choosing spot treatments instead of blanket applications or ordering biologics early so they are ready on time.
Advantages to the grower are the reason for adopting any new tech, not just imaging. These systems are reducing costs for labor and pesticides and increasing crop quality and consistency.
Real-time Coverage Enables Early Action
Maverick from “Top Gun” isn’t the only one who can feel the need for speed. A major benefit to the operator is the availability of real-time data and actionable intel. Early detection means early resolution.
“We can reduce waste in the industry,” says Greenberg. “We don’t have to accept the status quo. An opportunity to increase profitability is to grow more with the same inputs.”
Less waste means more marketable products from the same space in the same timeframe. Computer vision systems can detect issues two to three days earlier than a human scout is likely to notice, says Manotas. The images are at the granular level and can be analyzed for each visible leaf, which even trained, experienced workers find difficult.
Revol Greens CEO Michael Wainscott says they are still in the process of refining how they use imaging systems but are already seeing benefits.
“We are detecting tip burn three days earlier than with just grower eyesight,” he says. “We can also go back and look at the environmental conditions that created the problem.”
Tip burn or chlorosis on lettuce can be found, identified, and treated in time to save the crop instead of losing it. Or, give your team a warning of a shortfall.
“Letting you know you have a problem with your forecast is huge,” says Karin Tifft, Horticulturist at IUNU.
Interestingly, real-time imaging tagged to the crop can let you look back in time to find the cause of an issue.
“The grower can take those same plants and follow them back for the entire life of the crop, watching the damage increase daily, and can see what the climate was like. For example, how long was that temperature spike?” Tifft says. “By knowing where and what the problem is, the grower can train themselves and our AI on earlier symptoms, such that when the temperature spikes again, the grower can make corrections proactively, before symptoms, avoiding the loss next time.”
Early detection allows growers to be proactive instead of reactive. Fewer pesticides are needed, biologics can be ordered early, and blanket treatments can often be avoided — no need to nuke the whole place and then replace all your biocontrols. That translates to reduced input costs and reduced crop loss. Pesticide costs are often reduced by 20% to 30%, says Manotas, and customers see about a 25% reduction in integrated pest management (IPM) costs as well.
Accurate and Consistent Total Coverage
Complete coverage is also a significant advantage. With overhead cameras or, in the case of tomato growers, imaging systems on trollies, growers can cover the entire crop if desired — something impossible for larger operations with human scouts.
“You can’t watch 32 acres of leafy greens,” says Wainscott. “Cameras are the best way to do it.”
The granularity of the images and their consistency means that plant growth can be monitored, and the rate at which plants grow can be analyzed as well, explains Greenberg. For example, a growth slowdown caused by Pythium can be detected days before humans would notice — if they did.
Labor Issues
More than 80% of growers across all crop segments struggled with recruitment, retention, and/or training employees, according to IUNU’s State of CEA 2023 report.
Skilled human scouts often have a high turnover rate, and training takes time and money. A skilled scout can take years to develop the sixth sense needed to recognize and diagnose issues efficiently, says Tifft. Sometimes, those people don’t want to stay in the role or might not want to move to a new location. With computer vision and AI to recognize issues early, growers can harness highly accurate information and efficiently act on it, whether by themselves or by sending the data to a consultant, potentially solving a labor shortfall.
Computer vision and AI systems will be standard on all Revol Greens new builds, says Wainscott.
“Either double or triple your grower and assistant grower count or move to automation and AI,” he says.
For now, a human is still needed in the loop, says Manotas. While the system will provide actionable intelligence and likely diagnosis, confirmation by a trained IPM person is still recommended, just like a doctor will still order a blood test to confirm what they see. Over time, the system gets smarter (machine learning) and will provide results: whitefly infestation in houses 2, 3, and 5 between rows X and Y, for example. He projects that in the next several years, systems will evolve to detect, identify, and treat problems independently.
Greenberg sees a shift from the current greenhouse climate-based open loop to a plant-centric feedback loop with the strategy driven by the grower but implemented by the system.
With major growers widely adopting the technology, and positive results for the bottom line, computer imaging and AI are likely to become the standard in the near future.