Artificial intelligence is finding a new role in agriculture: helping farmers decide when their crops will be ready to harvest.
From apples and strawberries to blueberries and tomatoes, new AI-powered systems are using cameras, drones, smartphones and weather data to monitor crops and predict when fruit will reach the right stage for picking.
For farmers, getting that timing right can make a significant difference. Harvest too early and the crop may not be ready. Wait too long and fruit can deteriorate, while labour and other costs can increase.
AI watches crops as they grow
One approach involves installing cameras on tractors and other farm vehicles.
As the vehicle moves through an orchard or field, cameras capture thousands of images of plants. Artificial intelligence can then analyse those images to identify flowers, developing fruit and other indicators of crop growth.
US apple producer Okanagan Specialty Fruits, for example, has been testing technology developed by Canadian agricultural technology company Vivid Machines.
The system analyses images captured as farm machinery travels through orchards and can provide estimates of crop volumes and potential harvest dates.
This gives growers another source of information when planning their harvest.
Why predicting harvests matters
Harvesting fruit requires careful coordination.
Farmers may need to arrange seasonal workers, transportation, storage and buyers before picking begins. Weather can complicate those plans further.
Some fruits also have extremely short harvesting windows.
Berries such as strawberries can deteriorate quickly once they are ready, making accurate forecasts particularly valuable.
UK agricultural technology company FruitCast uses AI to forecast harvests for crops including strawberries, raspberries, blackberries, blueberries and tomatoes.
Its technology analyses images collected using drones, smartphones or cameras mounted on agricultural vehicles.
Weather data can improve predictions
Looking at fruit alone is not always enough.
Temperature, rainfall, irrigation and other environmental conditions can influence how quickly crops develop.
AI systems can combine visual information from plants with weather and irrigation data to improve their forecasts.
This could become increasingly useful as farmers deal with unpredictable weather conditions.
Extreme heat, drought and unusual rainfall patterns can change growing conditions and make traditional harvest schedules less reliable.
Smartphones and affordable drones could help
Advanced agricultural technology does not necessarily require expensive machinery.
Researchers are developing systems capable of analysing photographs taken using smartphones or relatively inexpensive drones.
At North Carolina State University, researchers have worked on technology that can automatically count blueberries appearing in smartphone images.
Researchers at the University of Florida have also explored crop-counting technology using images collected by relatively low-cost drones.
These approaches could eventually contribute to systems that estimate both how much fruit will be produced and when it will be ready.
Technology can even look inside fruit
Scientists are also experimenting with more advanced methods for determining ripeness.
Researchers at Princeton University have explored using millimetre-wave sensing to examine characteristics associated with fruit ripeness without cutting the fruit open.
The technology uses high-frequency radio waves that can respond to characteristics such as water and sugar content.
Information from technologies like these could eventually be combined with AI models to provide farmers with increasingly detailed information about their crops.
AI still needs good farm data
Artificial intelligence is not automatically accurate simply because large amounts of technology are involved.
Agricultural conditions can differ significantly between farms. Soil, weather, crop varieties, irrigation methods and farming practices can all affect production.
Historical information from an individual farm can therefore be particularly valuable when training or improving forecasting systems.
This means an AI system designed for agriculture may perform better when it understands the conditions of the specific farm where it is being used.
Farmers will still make the final decision
AI harvest forecasting remains an emerging technology, and researchers and agricultural companies acknowledge that these systems are still developing.
Farmers also need evidence that investing in new technology will provide enough value to justify its cost.
There are additional questions around agricultural data. Some growers may be reluctant to share commercially sensitive information about irrigation, fertiliser use or production methods with technology providers.
For these reasons, AI is more likely to support farmers rather than completely replace their judgement.
Experienced growers understand their crops, land and local conditions. AI can provide another layer of information to help them make those decisions.
What could this mean for Rwanda?
This technology could eventually have interesting applications for Rwanda and other African agricultural markets.
Smartphones, cameras, drones, weather information and AI could potentially help farmers monitor crops, detect problems earlier, estimate yields and determine better harvesting periods.
For high-value crops, better forecasting could also help farmers coordinate workers, transportation and buyers while reducing unnecessary losses.
The biggest opportunity may not be replacing farmers’ knowledge with artificial intelligence.
Instead, it could be combining farmers’ experience with better data, giving them more information to decide when and how to harvest their crops.
As agricultural AI becomes cheaper and more accessible, technologies that are currently being tested on large farms could eventually become practical tools for farmers around the world.
