Lead Researcher: Sean Smukler
Summary: This project seeks to train models using a series of multispectral satellite images to detect the types of plants growing in each of the UBC farm plots over a 5 year period (2021-2025). The results were used to inform a pipeline for developing weed detection/prediction tools for commercial farm fields through the prairies.
Project Contact: Sean Smukler
Summary: TreeTrack is developing an AI-enabled UAV monitoring platform to automate seedling survival verification in reforestation and regenerative land management settings. Current monitoring methods rely on manual ground sampling of 2–5% of planted areas, which is labor-intensive, slow, and statistically limited. This project evaluates whether high-resolution UAV imagery combined with machine learning can deliver accurate, full-site survival assessments in a faster and more cost-effective manner.
The primary research questions are:
- Can UAV-based imagery achieve greater than 90% seedling detection and classification accuracy under real field conditions?
- Can automated geospatial processing produce standardized survival reports within 72 hours of data capture?
- How does full-site aerial monitoring compare with traditional manual sampling in terms of coverage, efficiency, and data reliability?
The methodology involves periodic UAV flights using TreeTrack’s DJI platform equipped with high-resolution RGB imaging. Imagery will be processed through a cloud-based AI pipeline that detects, counts, and classifies seedlings. Limited ground-truth sampling will be conducted for validation purposes only; no planting, soil disturbance, or crop manipulation will occur.
The project’s broader impact is to improve transparency and accountability in ecological restoration and carbon markets by reducing monitoring costs and increasing data fidelity. Knowledge mobilization includes sharing results with CENGN and UBC Farm stakeholders, providing demonstration sessions, and engaging students interested in applied agri-tech and remote sensing.