Real-time UAV-based Fruit Perception in Orchards

Fruit growers need timely information about their crops to plan work and prepare for harvest. My PhD research explores how drones and artificial intelligence can turn images of grape bunches and apples into useful information more quickly. I developed and tested methods to reduce the amount of image data that needs processing or transmission, recognize small fruit, and combine views from different sides of trees to estimate apple numbers per tree. In controlled tests, processing was fast enough to keep pace with a drone survey. However, fruit that is too small in the images or hidden behind leaves remains difficult to count. These findings bring drone monitoring closer to providing useful information during or shortly after a flight. In future, this could help growers inspect their crops more efficiently and make better decisions about crop management and harvesting.