Computer scientists at the University of Canterbury have developed artificial intelligence that peers through dense foliage to locate and measure hidden fruit. Professor Richard Green leads the UC Vision project, where Dr. Richie Ellingham created the AI and computer vision system designed specifically for horticultural applications.
The technology addresses a persistent challenge in fruit farming. Harvesters traditionally must part leaves manually to spot ripening fruit, a labor-intensive and time-consuming process. The new system automates this detection, identifying fruit obscured by branches and leaves while simultaneously measuring their size and tracking their location and growth over time.
The approach combines machine learning with advanced imaging techniques to see through the visual obstruction that natural foliage creates. By training the AI on large datasets of horticultural imagery, researchers enabled the system to recognize fruit characteristics even when partially or fully hidden from direct view.
The implications extend across multiple sectors within agriculture. Automated fruit detection could reduce harvesting time, lower labor costs, and improve efficiency in commercial orchards and greenhouse operations. The ability to track individual fruit growth patterns also provides farmers with data to optimize ripening schedules and harvest timing, potentially increasing yield quality and consistency.
UC Vision's work represents a practical application of computer vision technology that moves beyond laboratory demonstrations into real-world farming scenarios. Horticulture has historically lagged in automation compared to other agricultural sectors, making this development particularly relevant for modern farming operations facing labor shortages and rising operational costs.
The researchers have not yet disclosed details about deployment timelines or whether commercial partnerships are in development. The technology's effectiveness across different fruit varieties and growing conditions requires further validation before widespread adoption becomes feasible.