Differential Digital Surface Models in Urban Environments

This Phase II investigates algorithms that create differential 3D models in near real-time to find differences in scenes, detect hidden motion, and compare current imagery with baselines. The approaches examined include very fast (near real-time) 3D reconstruction from multi-look imagery and related change detection techniques. A significant revision of our Structure from Motion (SFM) and Bundle Adjustment (BA) pipeline, BA4S, will be considered which will enable the system to fuse multiple/different flight orbits (i.e. different times and different trajectory specifications). The work also requires very precise calibration of the 3D models with consistent coordinates systems and differentiable 3D/DSM (Digital Surface Model) measurements. The approach also examines refinements to the approach such as superresolution in 3D reconstruction for refining the differencing process.