When Professor Kyeong-Hwan Lee watches a drone glide over an apple orchard in South Korea, he sees more than just pretty pictures. He sees the future of farming. Lee and his team at Chonnam National University have built an AI system that combines bird's-eye aerial views from drones with ground-level laser scans from robots to create incredibly detailed digital maps of orchards. The maps are so precise they can pin down the location of every tree trunk down to a few centimeters.
The problem the researchers solved sounds simple but is actually tricky. Drones can photograph orchards from above, but their cameras cannot see through dense tree canopies to what lies beneath. Meanwhile, ground robots equipped with laser scanners can capture fine details about individual trees, but they tend to lose their sense of direction over long distances because their satellite signals get blocked by all those leaves and branches. Lee's system solves both issues by having the drone images serve as a fixed reference point that keeps the ground robot on track.
"Our system utilizes deep learning-based cross-modal alignment to connect these two sources of information, identifying the same tree rows, canopy patterns and open spaces in aerial images and in the robot's laser measurements," Lee explains. "This way, the aerial map can be used as a reliable geographic reference, placing the ground robot's detailed measurements accurately within the larger orchard map."
In field tests, the system tracked a robot traveling about 1.3 kilometers through an apple orchard and maintained localization accuracy on the order of a few centimeters. It also proved robust to seasonal changes, meaning it works whether the trees are leafy in summer or bare in winter. The researchers published their findings in the journal Artificial Intelligence in Agriculture.
So what does this mean for the rest of us? These ultra-precise maps give farmers powerful new tools. They can estimate how much fruit an orchard will produce before harvest, spot trees that are sick or struggling, and guide robots to spray exactly where fertilizer or pesticide is needed without wasting it on empty rows. The system can also be loaded onto embedded computing devices, meaning the robots could eventually run these calculations in real time while working the fields.
Precision agriculture is all about using data to farm smarter, not harder. By combining the big picture view with ground-level detail, Lee's team has given growers a digital backbone that could make orchards more productive, more sustainable, and easier to manage. And that is something worth celebrating.