AI tools are helping fruit growers predict the best time to harvest
Camera systems, weather data and machine learning are being used to forecast when crops will be ready, although growers’ own experience remains central to harvest decisions.
By BBC News
Fruit growers are increasingly being offered artificial intelligence tools designed to predict when crops will be ready to harvest, as technology companies try to improve one of farming’s most difficult planning decisions.
The systems combine images of developing fruit with information such as weather, irrigation and previous crop performance to estimate future yields and harvest timing.
One example highlighted by the BBC is technology from Canadian company Vivid Machines, whose camera systems can be mounted on farm vehicles and used to analyse orchards as equipment moves between rows.
The company says its system can monitor individual trees from early growth through to harvest, recording information including bud numbers, flowers, fruit counts, size and maturity.
That data can then be used to support decisions about thinning, labour requirements, expected yields and when different parts of an orchard are likely to be ready for picking.
Vivid Machines says the technology is designed to supplement growers’ decision-making rather than operate independently of them.
A similar approach is being developed in the UK by FruitCast, which provides AI-based forecasting and farm management software to fruit growers.
Its system analyses images of crops captured using methods including farm vehicles, smartphones and other camera equipment, combining those observations with additional information such as weather and irrigation conditions.
FruitCast currently focuses particularly on soft fruit. The company says its forecasting technology is used across a significant share of the UK strawberry sector and is being extended across other crops.
Its forecasts are designed to estimate both the quantity of fruit likely to be picked and when it will become ready.
That timing can be particularly important for berries because the window between ripening and deterioration can be comparatively short.
Raymond Martin, FruitCast’s founder and chief operating officer, told the BBC that experienced growers generally know how their crops are progressing, but the challenge becomes greater when farms extend across large areas or include a mixture of indoor and outdoor production.
The company says its technology allows observations to be made at a scale that would be difficult to achieve through manual crop inspections alone.
FruitCast reports that its forecasts typically come within 10% of the volume eventually harvested when predicting one week ahead, equivalent to about 90% accuracy.
At three weeks ahead, it says forecasts are usually within 17% of the final picked volume. These figures are company-reported performance claims rather than independently established guarantees of results on every farm.
Forecasting remains affected by conditions in individual fields and orchards.
Vivid Machines, for example, says useful predictions depend on collecting detailed information from the farm itself rather than relying on general assumptions about a particular variety of fruit.
The companies therefore position AI as an additional source of information for growers rather than a replacement for agricultural experience.
For farmers, more accurate forecasts could help with decisions including how many pickers will be required, when packing and storage capacity will be needed and how much produce can be supplied to customers.
But the technology remains dependent on the quality of the information collected and on growers interpreting those forecasts alongside conditions on the ground.
For the foreseeable future, the decision over exactly when to harvest is therefore likely to remain a combination of data, technology and human judgement.