Researchers demonstrate drone-based AI vision system for automated tree pruning using stereo cameras

A research team published a system that mounts two cameras (stereo vision) on a drone to automatically detect radiata pine branches needing pruning. The system uses machine learning models (YOLOv8, YOLOv9, Mask R-CNN) trained on 71 image pairs to identify branches and calculate their exact distance from the drone using depth estimation—a technique that measures how far away objects are by comparing two camera angles.

The system tested six different depth-calculation methods and found that deep-learning approaches (neural networks trained on visual data) produced clearer, more reliable distance measurements than traditional computer-vision algorithms at working distances of 1–2 meters, making low-cost stereo camera setups viable for this task.

For forestry companies and agricultural equipment manufacturers, this work removes a major barrier to automated pruning: the ability to precisely locate branches in 3D space. Previously, drones lacked the perception needed to decide which branches to cut; now the path toward autonomous pruning equipment that reduces manual labor in tree maintenance becomes technically feasible.