Challenge
Autonomous operation of forestry vehicles can only be planned at scale if the health, species, and condition of individual trees is known and precisely located.
Solution
We built a drone-based solution that flies the forest ahead of harvesting to determine 3D surface structure and localize trees with their characterization (diameter, species, bend), fusing RGB camera and LiDAR data with available national data to produce an enriched operational map.
Results
- Data collection platform for surface reconstruction and object recognition
- Semantic segmentation of the forest
- Data fusion from LiDAR and cameras
Technology
Python · Open3D · OpenCV · PyTorch · MM Segmentation
Powered by DAC Digital