3DWAR: 3D Wide Area Reconstruction

The 3D Wide Area Reconstruction (3DWAR) effort brings the operational benefits of superior 3D reconstructions to the warfighter. Superior performance is achieved by combining super-resolution (SR) and computational stereo processing to reconstruct 3D models of the combat environment with voxel size of one half the imagery ground sample distance. If input data has sufficient 3D visual effects to support stereo reconstruction, then those effects are also sufficient to derail any SR processing that does not explicitly account for them. We extend our multi-view Accumulation-Based Modeling (ABM) computational stereo approach to tailor it to the operational problem. ABM is naturally compatible with wide field of view and rolling shutter cameras due to its unique approaches to evidence fusion and interpretation. Phase 1 establishes the feasibility of the proposed technical approach with three spirals of theoretic and empirical analysis. Together, these spirals lay a solid foundation for Phase 2 by providing proof of concept implementations and quantitative performance analyses of the approach. Phase 2 will augment and fully develop the most promising variations, and demonstrate an integrated engineering prototype on operationally relevant imagery. The proposed work leverages our past experience in computational stereo and super-resolution, as well as our existing stereo capabilities.