Mark Venables spoke to James French, Director of Systems & Sensors at TOMRA, to understand how the company’s AI capabilities are setting new standards in produce processing while helping the industry respond to the pressures of population growth and environmental constraints.
Artificial intelligence (AI) is making waves across multiple industries, and agriculture is no exception. AI addresses longstanding challenges in fresh fruit processing and packing, such as product variability, unpredictable demand, and high labour costs. By integrating AI-powered technology into sorting systems, TOMRA is transforming the industry, enhancing quality, efficiency, and profitability.
The demand for reliable, high-quality produce has never been more urgent. “AI plays a critical role in enhancing agricultural yield by reducing food production losses and waste,” James French, Director of Systems & Sensors at TOMRA, says. “With rising global populations and increased consumption, we must optimise our agricultural land use. AI can help us make informed decisions in response to changing environmental factors, like climate change and geopolitical influences, which impact the consistency of food production.”
TOMRA’s AI-based systems also improve product quality and uniformity, enabling fresh, unprocessed fruits and vegetables to compete with highly processed alternatives. “Our primary focus is on ‘whole foods,'” French continues. “These are the fruits and vegetables that customers buy without further processing, though we also work with some processed items, like French fries. AI has helped us improve the quality and appeal of these whole foods.”
Agriculture might not be as data rich as other sectors, but the potential for leveraging data remains vast. “Post-harvest, when produce enters processing facilities, data volumes increase significantly through cameras and sensors monitoring each item,” French explains. TOMRA’s sorting systems use this data for immediate, real-time decisions that impact efficiency and waste reduction. However, French highlights a lost opportunity, as much of this data is discarded after processing.
AI systems like TOMRA’s offer a more sustainable approach to data by connecting the dots along the supply chain and retaining insights for long-term improvement. “Traceability efforts are now about connecting existing data rather than gathering new information,” French says. “We are focused on retaining and using data that is often discarded after initial processing. This enables us to offer robust traceability without additional data collection, making it cost-effective and practical.”
The role of Lucai: enhancing defect detection for industry giants
One of TOMRA’s flagship AI technologies, Lucai, has redefined defect detection and classification in the fresh produce sector. Built on extensive datasets, Lucai’s deep learning algorithms enable it to identify and classify even the smallest defects across various fruits. The system’s adaptability and ability to update grading parameters based on market demands ensure that it remains relevant and effective in any environment. French highlights its efficiency: “We use AI to make sorting decisions, allowing customers to identify attributes of their products with unparalleled accuracy. Lucai can detect tiny blemishes on blueberries, splits in apples, or decay on cherries, improving quality and reducing waste.”
The impact of Lucai is already evident at FreshCo, a significant player in New Zealand’s agricultural sector. Robin Mudgeway, FreshCo’s Technology & Machinery Manager, underscores the value of AI in simplifying operations: “With AI, we are not constantly having to fine-tune the system or deal with misidentifications. It eliminates the issue of defects around the stem area, which used to require manual mapping and adjustments.”
Glen Kaunds, Senior Application Engineer at TOMRA, also notes Lucai’s seamless functionality: “Spectrum’s AI integration is plug-and-play. You turn it on, and the pre-trained models start grading with high accuracy, identifying nearly all defects without complex adjustments.”
The integration of Lucai into FreshCo’s systems has improved efficiency and precision, with Mudgeway adding, “AI handles grading for us, allowing production to continue at speed without the need for manual sorting on the line. Our return on investment with AI has been undeniable, and we’re likely to see a full return within two years.”
Maintaining quality in a high-pressure environment
Prima Frutta Packing, a top supplier of California cherries and apples, attests to Lucai’s game-changing capabilities in the United States. For President Tim Sambado, maintaining quality is the ultimate priority. “Our customers expect a premium product, and we work hard to meet their needs daily,” he says. “We’re obsessed with using the best technology available, and data drives every part of our operation. Lucai has enabled Prima Frutta to improve sorting accuracy and enhance operator efficiency, increasing throughput and consistency.”
Sambado highlights the advantages of Lucai over traditional methods: “Our investment in Lucai had three main goals: improving accuracy, simplifying the system for our operators, and enhancing our ability to detect specific defects that were difficult to catch. Lucai has met our expectations, and as we continue to refine it, the system is learning to detect specialised defects like powdery mildew and insect bites.”
For Sambado, Lucai’s intuitive user interface has been especially beneficial in a fast-paced environment. “The new user interface that we installed with TOMRA allows us to make adjustments and immediately see the impact on grading,” he notes. “When you’re running 40,000 cherries a minute, that instant feedback is invaluable. With Lucai, we’ve increased production without expanding our labour force, enabling us to keep up with the intense pace of California’s cherry harvest.”
The broader impact of AI in agricultural sorting
AI is not just streamlining processes; it is bringing new levels of adaptability to an industry that relies heavily on variable raw materials. “Unlike in manufacturing, where products are more uniform, agricultural items are inherently variable. Every apple or cherry is unique, as are their imperfections,” French explains. “AI excels at interpreting this variability because it learns from large volumes of data, enabling it to make robust, reliable decisions.”
With deep learning techniques developed in-house, TOMRA’s AI is designed for real-time decision-making and high-volume operations, avoiding recalibration. “Our systems can remove unsatisfactory products from processing lines based on AI insights and even adapt between different fruit varieties without additional calibration,” French adds.
This flexibility has proven essential in managing sudden changes in product quality due to weather events. At FreshCo, for example, AI was instrumental in handling the aftermath of a severe storm that increased stem splits in Royal Gala apples. “AI handled these defects effortlessly without slowing down production, which keeps throughput at an optimal level,” Mudgeway reports.
Challenges and the path ahead
Despite its potential, implementing AI in agricultural settings comes with its own set of challenges. One significant hurdle has been the data demands of AI development. “Developing our blueberry models required over 100,000 individual scans,” French notes. “There is also a need for specialised talent, like data scientists, which can be difficult to secure. Educating teams and customers about AI’s capabilities and limitations has also been critical, especially to dispel misconceptions arising from media hype.”
AI, however, has introduced efficiencies that balance these demands. “In some ways, AI has simplified our processes because it replaces multiple algorithmic steps with a single, trained network,” French observes. “This simplification reduces the need for constant recalibration, freeing up resources and allowing for more streamlined operations.”
French is optimistic about the future of AI in agriculture, particularly as more touchpoints in the agricultural process become interconnected. “In the near term, we expect to see pre-harvest yield forecasting, sorting, and shelf-life estimation becoming more interconnected, optimising decision-making across the value chain and aligning output with consumer demand.” Yet, he acknowledges that realising an end-to-end AI solution for agriculture is complex, and achieving full integration across the supply chain remains a long-term goal. Some companies, like New Zealand’s Zespri, are already paving the way by managing the entire value chain from farm to consumer, proving that holistic AI-driven solutions are possible.
In addition to developing its AI algorithms in-house, TOMRA collaborates with external AI experts such as Faktion to accelerate innovation and focus on real-world applications. “Our approach focuses on practical, real-world applications,” French says. “We work with partners who share our pace and objectives, avoiding the delays often associated with academic partnerships.”
In the case of Prima Frutta, this partnership-based approach has brought tangible results, as Sambado highlights: “Our partnership with TOMRA has been crucial. We encouraged TOMRA to move into artificial intelligence for the past few years, and they listened. Developing and refining the AI has been smooth, and TOMRA remains committed to improving it. Every day, we encounter new challenges in the cherry business, and it’s exciting to see how these systems can adapt and improve.”
AI for the future of agriculture
As AI continues to reshape agriculture, companies like TOMRA set the stage for a future where efficiency, sustainability, and profitability are deeply intertwined. For both FreshCo and Prima Frutta, TOMRA’s AI-driven systems are more than just a technological upgrade; they represent a fundamental shift in how fresh produce is processed and delivered to consumers.
Sambado aptly concludes, “In this business, the key isn’t just how you start, but how you continue and finish. Every day brings new challenges, and seeing how AI can adapt and improve is exciting. We have only scratched the surface of what’s possible with AI-driven sorting.”
The future of fresh produce lies in harnessing data to tackle the complex challenges facing agriculture. With AI at the helm, TOMRA’s technology is raising industry standards and paving the way for a more intelligent, more resilient food system.
