Our Research Themes
ASIL’s research is organized around four interconnected themes. Each theme combines multiple data modalities — imagery, LiDAR, multispectral and radar sensing — with advanced analytical methods (deep learning, computer vision, photogrammetry, point cloud analytics) to produce actionable geospatial intelligence for partner communities and institutions.
Geospatial AI

Developing deep learning and computer vision methods for geospatial data — including image classification, object detection, change detection, and multi-modal data fusion. The lab has contributed to:

  • Multimodal building footprint and roof structure extraction from aerial imagery, orthophotos, and LiDAR
  • Cross-modal camera-LiDAR registration using feature-based matching and graph neural networks
  • Sensor fusion for UAV-based imaging and pansharpening
  • Off-nadir change and damage detection using deep learning

Recent publications in IEEE JSTARS, ISPRS JPRS, Remote Sensing, and Canadian Journal of Remote Sensing demonstrate the lab’s contributions in this area.

Urban Digital Twins

Building accurate, scalable 3D representations of cities at Levels of Detail 2 and 3 — including buildings, facades, windows, doors, and other components — for downstream applications in energy modelling, planning, and resilience analysis.

Current and recent work includes:

  • LoD2 city modelling using hybrid data-driven and geometric approaches, demonstrated across New Brunswick urban areas
  • LoD3 facade element reconstruction from multi-view imagery and mobile scanning systems (awarded Best Paper, CJRS 2024)
  • Building digital twins for thermal and energy modelling, including ongoing collaborations with Natural Resources Canada
  • Maritime Digital Twin Project for port resilience (ACOA-funded, with the University of New Brunswick Marine Sciences group)
Remote Sensing for Climate Resilience

Applying remote sensing to climate adaptation and resilience problems — particularly in the Canadian context, with applications in flood mapping, urban heat analytics, and dike infrastructure planning.

Specific contributions include:

  • SAR-based urban flood detection using TerraSAR-X and Sentinel-1 imagery
  • Super-resolution flood mapping with deep learning (ESRGAN-based methods)
  • Urban heat island projection to 2050, integrating climate change and urban development scenarios
  • First-floor height estimation for building-level flood damage models, using vehicle-mounted imagery and airborne LiDAR
Disaster Management & Damage Assessment

Rapid, accurate assessment of building damage and disaster impact from very-high-resolution imagery — supporting first responders, government agencies, and insurance assessors.

Areas of active and published work:

  • Post-event damage assessment using deep transfer learning on very-high-resolution imagery
  • Earthquake building damage detection combining textural and information-theoretic analysis (CJRS 2nd-best paper, 2021)
  • Real-time disaster mapping and dashboarding with Natural Resources Canada
  • Concrete infrastructure inspection through AI-guided crack detection using smartphone sensors
Get Involved

Each of these themes welcomes new contributors. If you’re a prospective graduate student, postdoctoral fellow, undergraduate researcher, or external collaborator interested in working on questions like these, see the Join Us page for current opportunities and application details.

For specific project-level questions or collaboration inquiries, contact Dr. Shabnam Jabari.