A new AI system that counts flowers on fruit trees is helping farmers to predict harvest sizes months in advance, in turn making crop yields more profitable.
The system uses images taken with a standard smartphone to estimate the number of flowers on a fruit tree by recognising patterns and features, such as the edges and shapes of petals, even when they overlap or are partially obscured.
When tested on peach orchards in Spain, the AI predicted flower counts with a 90% accuracy - a big improvement on current manual methods used by farmers, such as counting by eye or taking area samples, which can have error rates of 30-50%.
By providing more precise yield forecasts up to six months before harvest, the system could help farmers, including growers in the UK, to optimise water use, allocate human and economic resources more efficiently, and better plan harvesting and distribution logistics.
It’s estimated that agriculture uses 65% of the world’s fresh water, almost half of which is wasted, whilst 45% of fruit and vegetables produced for human consumption are lost or wasted every year across the global supply chain.
Researchers from the National Robotarium, the UK’s centre for AI based in Edinburgh, will validate the project's predictions against the actual peach harvest in September 2024.
If proven effective, they believe the approach could be adapted for other important crops like apples, pears and cherries, benefiting growers in the UK.
Fernando Auat Cheein, associate professor in autonomous systems at the robotarium, said while the research focused on peaches, the principles behind this technology could be applied to UK crops.
“In countries all across the world, farmers often rely on manual methods to estimate their yields, which can have a significant error margin," he explained.
"By leveraging the power of AI and the convenience of technology like smartphones, our approach integrates with traditional farming practices, making it easier for growers to embrace from innovative solutions without overhauling their existing methods."
Throughout the project, researchers worked closely with peach farmers in Spain to understand their challenges and ensure the system met their requirements.
The farmers appreciated the simplicity and accuracy of the flower counting AI, noting its potential to help them make more informed decisions about crop management, such as targeted pruning and herbicide application.
Mr Cheein said: "By focusing their efforts on areas of the farm that are expected to yield the most fruit, farmers can optimise resources, reduce their environmental footprint, and maximise both the quantity and quality of their harvest."
