ELEV8
Status: Submitted
Rapid Digital Elevation Mapping with Event Cameras to reconstruct dynamic mountain environments in near-real-time.
ELEV8 â Rapid Digital Elevation Mapping with Event Cameras â develops a new generation of Digital Elevation Models (DEMs) for rapidly changing mountain environments.
We combine event-based vision sensors, UAV platforms, and real-time onboard computing to move from slow, light-sensitive terrain mapping to robust, nearâreal-time elevation reconstructionâeven in shadows, snow, and high-contrast conditions.
Scientific Goals
- High-frequency updates: Enable near real-time DEM updates for dynamic landscapes like glaciers and landslides.
- Efficiency: Significantly reduce processing time and computational cost compared to classical photogrammetry.
- Accuracy: Achieve decimetric vertical accuracy comparable to SfM and LiDAR.
- Dynamic Observation: Shift DEMs from static products to continuously observable variables.
Methodology
- Hybrid Event + RGB Approach: UAVs carry both sensors. Event data are used to constrain and accelerate feature matching, resulting in a 10Ă reduction in matching complexity.
- Pure Event-based Reconstruction: Elevation is inferred directly from asynchronous event streams using UAV motion and temporal super-resolution. FPGA-based onboard processing allows for DEM generation during flight.
Key Innovations
- Strengths from Weaknesses: Especially through the use of event cameras, converting traditional sampling weaknesses into strengths.
- Dynamic Signal Exploitation: Using shadows and contrast changes as useful signals rather than noise.
- Temporal Super-Resolution: Using UAV motion to create subpixel âvirtual resolution.â
- Efficiency: Real-time processing with low-latency, energy-efficient hardware.