← Back to Research
Research Note · Computer Vision & Aerospace
UAV Disaster Mapping
Published: Aug 2026•8 min read•Author: Chetraj Jaishi
UAVPhotogrammetryComputer VisionDisaster Relief
Abstract: Rapid post-disaster reconnaissance using aerial photogrammetry and semantic segmentation in mountainous topography.
Following seismic and landslide events in mountainous terrain like Nepal, access roads are immediately severed and traditional ground survey teams are endangered. Commercial satellites face cloud occlusion and multi-day revisit latency.
We investigate a modular UAV surveying framework utilizing terrain-following flight plans generated from SRTM elevation models, coupled with lightweight photogrammetry pipelines.
Captured high-resolution oblique photography is synthesized into georeferenced orthomosaics and digital surface models (DSMs) on portable rugged laptops in the field within 25 minutes of landing.
A fine-tuned semantic segmentation model classifies road fissures, debris flow corridors, and structurally compromised dwellings, exporting GeoJSON vectors compatible with first-responder GIS viewers.
Key Research Takeaways
- Terrain-following waypoints are mandatory in steep valleys to maintain consistent ground sampling distance (GSD).
- Field-deployable edge photogrammetry removes reliance on high-bandwidth cellular networks in disaster zones.
- Accurate elevation models allow calculating debris volume to estimate required heavy machinery clearing hours.