September 2020 - 30

MANAGEMENT
Using AI to Maximize
Yield and Efficiency
How can new technology help growers reduce
shrink and optimize space, resulting in an
increase in food accessibility?
BY CARL SILVERBERG
A
s the United States continues to grapple with COVID-19, existing
challenges within the food industry have become even more
apparent. More than $160 billion worth of food is wasted each
year; over 20 million people in America live in a food desert and lack
access to quality, nutritious food; and the workforce that is most crucial to
our domestic food production is aging - 65% of growers are above the
age of 55.
Artificial Intelligence (AI), the ability of technology to replicate human
sensing and thinking, is beginning to play an important role in the food
industry. Machine vision - the ability of machines to interpret visual
images and learn from them - is an application of AI. Also known as
computer vision, it will be impactful in reducing food waste, increasing
food accessibility and augmenting skilled labor across the value chain.
By leveraging computer vision in controlled environment agriculture
(CEA), growers can detect and address issues ahead of time, predict yields
and reduce shrink, as well as optimize space and resources to maximize
output and improve recipes.
AI FOR PEST AND DISEASE DETECTION
AI can help growers detect problems with crops throughout the
growing cycle. The treatment or removal of diseased crops is required
in any operation, outdoor or CEA. When a grower sees a sticky card
covered in pests or notices an anomaly in a crop such as damage, spots,
chlorosis or some other malady, they adjust controls or remove plants to
address the issue.
Unfortunately, manual pest and disease detection is only as accurate as
the human eye and memory, and takes hours of time each week.
Today, there are AI programs that provide disease and pest detection
and alerts on regular, frequent intervals. iUNU, a Seattle-based company,
operates an optical scanner system called LUNA that continuously
collects imagery in greenhouses and analyzes visual data from growth
30 SEPTEMBER 2020 GPNMAG.COM
cycles. The software is able to understand what healthy crops look like
and notify growers when problems appear.
As the system sees more images, it learns. Not only can it detect
an issue, but it can also diagnose the problem - whether it is thrips,
aphids, powdery mildew or other potential damage. It can also help
provide recommendations and record information to improve recipes in
future cycles. Cameras pick up details and day to day changes unseen
by the human eye, so diseases are flagged faster, giving growers the
time they need to find and fix problems.
When growers trust AI with pest and disease detection, they also
become more efficient with their time, eliminating hours spent on " crop
walks " checking on plants, and reducing possible transmission of such
pests between facilities.
AI FOR YIELD PREDICTION
AI can also drastically improve yield predictions. Accurate forecasting
is a crucial component of running a farm or greenhouse, as it directly
affects the ability to presell overage. If there's a surplus of product and
not enough demand, profits shrink and food waste increases.
Manual yield prediction is often imprecise and inaccurate, despite
notebooks full of recipes, Excel spreadsheets, and years of knowledge
and experience. Although growers have refined the ability to offer
highly calculated guesses based on prior crops and similar conditions,
the margin of error is significant. Halfway through a growth cycle, yield
predictions can be overestimated or underestimated by up to 30%.
Improved yield prediction is now possible due to predictive technology
and AI. Just as a machine can be taught to detect pests and pathogens
more accurately with more data, machine learning software can
remember yields from prior harvests, and improve forecasting accuracy
over time by making visual comparisons to prior growth cycles.
" LUNA's yield forecasting changes the game for growers, " says Adam
http://www.GPNMAG.COM
September 2020

Table of Contents for the Digital Edition of September 2020
