Overview
Remote Sensing systems—deployed on satellites, airborne drones, and field spectrometers—measure reflected and emitted radiation across hundreds of narrow spectral bands. The Remote Sensing research team at Vision Lab formulates state-of-the-art computational algorithms to extract physical, chemical, and biological properties of our planet from complex spectral datasets.
By integrating physical radiative transfer models with cutting-edge machine learning, we turn massive spatial-spectral satellite data into actionable insights for climate resilience, agriculture, and urban planning.
Methodological Strengths
Spectral Unmixing & Super-Resolution
In remote sensing images, a single pixel often covers a mixed surface area containing various materials (e.g., vegetation, water, asphalt, soil). We develop linear and non-linear spectral unmixing methodologies that identify constituent endmembers and estimate their fractional abundance within individual pixels.
Multi-Modal Data Fusion
Combining diverse sensing technologies enhances spatial, temporal, and spectral resolutions. Our fusion algorithms seamlessly integrate:
- High-spatial-resolution RGB and multi-spectral photography
- Continuous high-spectral-resolution hyperspectral imagery
- 3D structural elevation data from airborne LiDAR and Synthetic Aperture Radar (SAR)
Environmental & Societal Impact
- Precision Agriculture: Monitoring crop stress, nitrogen content, water deficit, and early fungal infections to optimize resource usage and boost agricultural yield.
- Forestry & Biodiversity Conservation: Mapping individual tree species diversity, canopy biomass, and carbon sequestration capacity across natural reserves.
- Disaster Response & Urban Monitoring: Tracking flood extents, wild-fire burn scars, urban heat island effects, and illegal industrial emissions.