🌳
Biodiversity Credit Mapping
Biodiversity credit schemes need defensible, per-tree measurements across large sites. Manual field counts could not keep up, so I turned to drone data.
I segmented drone point clouds and multispectral orthomosaics with Detectron2 / Mask-RCNN to isolate individual trees, then modelled tree properties and surface/canopy structure with scikit-learn.
- Delivered individual-tree delineation instead of coarse plot estimates.
- Linked canopy models to biodiversity indicators for credit accounting.
Developed while helping a large NGO in Australia with their Australian Carbon Credit Unit (ACCU) work.