University of Canterbury research is helping growers forecast harvests more accurately and opening the way for increasingly precise automation.
This content was prepared by the University of Canterbury and is being published by The New Zealand Herald as Advertorial.
What if growers could know exactly how much fruit is on their trees and vines, even when much of it is hidden behind leaves?
Researchers at Te Whare Wānanga o Waitaha | University of Canterbury (UC) are turning that possibility into reality by developing artificial intelligence (AI) technology that can detect and measure concealed fruit, helping growers forecast harvests more accurately.
In testing, AI developed by the UC Vision team has produced fruit counts within about 2-3% of the correct figure. Traditional manual sampling, still widely used to forecast harvests, can be inaccurate by up to 23%.
For large growers, that difference could be worth millions of dollars.
Led by UC Computer Science Professor Richard Green, the team has developed a first-of-its-kind system that uses cameras, lighting and artificial intelligence to digitally remove foliage and create a complete 3D model of each plant.
The technology can identify individual apples, cherries or grapes, show which branch or cane they are attached to, and calculate their size, volume and surface area. Repeated scans can also track how individual fruit grows over time.
“You can’t automate what you can’t see. By creating a complete 3D model of each plant, including what is behind the leaves, we can open the door to automating almost any repetitive manual task associated with that plant,” Professor Green says.
“We now have technology with the potential to change how food is grown and harvested around the world.”
Accurate crop forecasting is critical for orchards and vineyards. Growers need to know how much fruit they are likely to produce, when it will be ready and what size distribution to expect.
Those forecasts shape decisions about workers, packaging, transport and storage.
“Getting that wrong can mean having too many or too few resources available at harvest time,” Professor Green says. “Better information could potentially save large growers millions of dollars and reduce waste across their operations.”
The detailed plant models could eventually guide autonomous machines carrying out labour-intensive tasks such as pruning, thinning, spraying and harvesting.
In a cherry orchard, for example, a machine could identify and pick only fruit that had reached the most valuable size, then return later for the remaining crop.
“Cherries that are only a few millimetres larger can be worth twice as much,” Professor Green says.
The team has collected 3D modelling data from 20 commercial farms to ensure the system works reliably in real orchard and vineyard conditions.
UC Vision researcher Dr Richie Ellingham says growers are already asking when they can start using the sampling tools.
That demand has driven the establishment of HoloCrop, a new UC spinout developing precision horticulture tools to provide more accurate data, reduce food-production waste and support future robotics.
The AI breakthrough builds on about 15 years of UC research, supported by more than $32 million in government investment in computer vision, artificial intelligence and agricultural robotics.