| Abstract |
Wildfires are increasing, yet current vegetation fuel assessments using manual surveys or satellite imagery are costly and low in resolution. This project investigates whether an autonomous drone-based computer vision system can more accurately classify forest fuel types and generate flammability heatmaps. A custom-built drone captured aerial images, which were stitched and analyzed using a trained machine learning model to classify vegetation. The system produced meter-level resolution heatmaps identifying high-risk fuel zones. Results showed significantly higher spatial detail and faster data collection compared to conventional methods, demonstrating a scalable and cost-effective solution for wildfire risk monitoring. |
