January/February 2024 - 22

TECHNOLOGY
the line into different grades to enable
hands-free production.
For example, if heavy rain or frost
has damaged so much of the incoming
fruit that only 40% of it is packable,
older technologies would not be accurate
enough to recover this as they would
mistakenly include too much damage, but
AI makes it possible. As well as recovering
a potentially disastrous harvest, this helps
keep customers satisfied at times when
the product might otherwise fail to make
the grade.
DEEP LEARNING IN ACTION
Deep learning is an AI method
which uses pre-trained models to teach
computers how to process data, such as
complex patterns in photos. The newly
launched Spectrim X series grading
platform is a great example of this.
Developed by a team of industryleading
scientists, engineers, researchers
and experts, Spectrim X integrates the
latest development of TOMRA's LUCAi
Deep Learning technology.
The Spect r im X ser ies grading
platform is equipped with LUCAi
Engine Software, computing hardware
and pre-trained models which achieve
unprecedented grading precision. More
packing operations can now be achieved
without manual intervention, targeting
thresholds more closely and all while
minimizing fruit loss.
Spectrim X assesses thousands of highresolution,
multi-channel fruit images
every second. Then it cross-references
what it sees with AI networks that have
been trained on tens of thousands of
fruits to make grading decisions to meet
precise market demands. This data has
been captured from TOMRA machines
across the globe and accurately hand
labeled by TOMRA's Data Science Team.
During 18 months of real-world testing
in the USA and New Zealand, sorting
and grading apples, Spectrim X showed
a significant leap forward in performance
compared to its predecessor.
AI also powers the new TOMRA Neon
to pre-grade blueberries. Automated
blueberry harvesting is faster and less
costly than manual harvesting but brings
challenges in the form of unwanted debris
and fruit clusters to fresh fruit processing
and packing lines.
22 JANUARY/ FEBRUARY 2 024
TOMRA Food's new-generation Spectrim X series leverages deep learning for sorting and grading precision of fruit.
To address these challenges, TOMRA
Neon accepts and rejects fruit before
transferring it directly onto TOMRA's
KATO260 optical sorter and grader. It
is by using AI that TOMRA Neon is able
to identify, differentiate and remove
unwanted clusters, undersize fruit and
unripe fruit. This optimizes optical grader
efficiency by removing more than 95% of
clusters and more than 90% of unwanted
green and red berries.
REASONS FOR OPTIMISM
AI's valuable capabilities in food
processing will become even more
important in the near future because of the
increasing challenges presented by global
food scarcity and disruptive weather
events - and because there's commercial
pressure to provide the highest-quality
ingredients and products even when input
product quality is variable.
AI can also help processors and
packhouses deal with other challenges,
such as providing products and services
at the lowest cost; meeting customers'
product specifications without over- or
under-delivering; and reducing or removing
the problems associated with recruiting,
training and retaining skilled labor.
The headaches associated with finding
enough people to do seasonal work, and
the ongoing challenges of managing a
workforce, can be addressed by automating
tasks previously done manually. It's no
secret that sorting, grading and packing
machines are faster, more accurate,
consistent, reliable and ultimately more
cost-efficient than humans.
This brings us back to the controversial
subject of AI replacing jobs. It's worth
remembering that many processing
plants, especially those not near large
population centers, struggle to employ
all the people they need - so rather
than performing tasks previously done
manually, automation often does what
otherwise could not be done.
Automation often handles tasks
people don't want to do because they
are boring, repetitive or strenuous. This
is undoubtedly true of manual sorting and
grading, and if workers are displaced from
these roles, often they are redeployed to
other tasks on the line, which are less
monotonous and add more value.
As we've seen here, however, reducing
labor dependence is just one of many
ways AI-driven automation benefits
processors and packhouses. AI is set
to play an increasingly important role
in meeting the needs and desires of
food consumers as the world's growing
middle classes increase demand for
healthy food.
Most importantly, AI will help meet
the challenge of feeding the world's vast
and growing population, which will
require more food production and less
food waste.

January/February 2024

Table of Contents for the Digital Edition of January/February 2024