High Resolution 3D Reconstruction from Wide-Area Video

Applied Science Innovations, Inc. in collaboration with Lockheed Martin Corporation proposes development of the Super-resolved Volumetric Imagery from Video Aerial Surveillance (SVIVAS) – a novel set of image processing algorithms to produce, with modern wide area video sensors, highly accurate 3D models of the area under surveillance, with voxel sizes one-half of the sensor fundamental ground sample distance or better. Emerging wide-area video sensors such as Autonomous Real-time Ground Ubiquitous Surveillance - Imaging System (ARGUS-IS) cover large areas at high resolution. New techniques of superresolution and 3D model reconstruction with observations from multiple perspectives will advance capabilities of such systems even further. The proposed SVIVAS will provide 3D reconstruction with resolution better than half the ground sample distance from aerial imagers similar to ARGUS-IS, while addressing computational efficiency, effect of rolling shutter, superresolution beyond limitations by pixel sampling and by the optics, distortion and perspective correction due to wide field of view (>60 degrees), and other critical issues specific to imagers similar to ARGUS-IS. Limited to feasibility studies, Phase I will identify computationally efficient methods for 3D reconstructions across wide field of view electro-optic sensors on fixed-wing flying platforms. In Phase II, computationally efficient 3D reconstruction algorithms will be implemented for wide-field-of-view system similar to ARGUS-IS.