AI tools help farmers forecast when fruit is ready to harvest

AI tools are helping some fruit growers estimate when crops will be ready to pick, but their forecasts depend on farm-specific data and are not a finished solution.

Okanagan Specialty Fruits, which has more than 1,250 acres of apple orchards in Washington State, is testing tractor-mounted cameras from Canada’s Vivid Machines. The system uses AI to identify buds, flowers and fruit, then provides crop estimates and harvest dates. Company representative Joel Carter says it can spot tiny flower buds, while forecast accuracy depends on the quality of historical farm data. He says Granny Smith apples have a harvest window of about three weeks, compared with only a few days for some berries.

UK-based FruitCast forecasts crops including strawberries, raspberries, blackberries, blueberries and tomatoes, using footage captured by drones, smartphones or cameras on farm vehicles. Co-founder Raymond Martin says the tools help assess crops across large farms. FruitCast says its forecasts are within 10% of picked volume one week ahead and within 17% three weeks ahead, and that it guarantees less than 20% error.

Angus Soft Fruits operations director Neill Finlayson said forecasting technology has made progress, but the industry is still some way from a fully integrated system. Researchers are also exploring millimetre waves to detect fruit ripeness. At Princeton University, Yasaman Ghasempour and her students have developed a detector that could be used by growers or shoppers.

Adoption remains a challenge. North Carolina State University’s Jing Zhang said growers must decide whether investing in new technology is worthwhile.