KITWARE INC — Department of Defense SBIR Phase I: ABSTRACT: Kitware proposes MAP-Tk, the open-source Motion imagery Aerial Photogrammetry T

KITWARE INC — SBIR Phase I award from Department of Defense.

Amount
$149,963
Agency
Department of Defense · Air Force
Program / Phase
SBIR · Phase I
Solicitation
2013.1
NAICS
Place of performance
NY
Period
2013-09-05 → 2014-06-06

Description

ABSTRACT: Kitware proposes MAP-Tk, the open-source Motion imagery Aerial Photogrammetry Toolkit, designed for optimized sparse bundle adjustment (SBA) of FMV and WAMI. Our proposed toolkit has a highly configurable and modular design supporting interchange of existing solutions for feature detection, description, correspondence, and SBA. Each module will build on existing open source software, including VXL and OpenCV. Kitware will also provide a custom solution for each module that is optimized for aerial video. In particular, Kitware will provide a GPU accelerated feature detector and descriptor using OpenCL. Kitware will also provide a feature correspondence module that replaces pairwise frame matching with sequential tracking assisted by a dynamic search radius around each feature point. Robust homography estimation will provide geometric constraints on the matching and these same homographies will guide the search for loop closure. Our sparse bundle adjustment module will select key frames in a temporal hierarchy to significantly reduce the problem size and rapidly obtain a coarse-scale solution before interpolating and optimizing the cameras of the remaining frames. It will also reduce the problem size by limiting the number of points in the reconstruction. MAP-Tk will support both archive-mode, for maximum accuracy, and streaming-mode, for real-time operation. BENEFIT: The primary benefit of the proposed work is a solution to problem of camera calibration from motion imagery that is far faster than the existing state-of-the-art. This solution will reduce reliance of intelligence, surveillance, and reconnaissance (ISR) systems on inertial navigation and global positioning, which are often unreliable or unavailable. Likewise, it will have the same benefit for commercial mapping applications that rely on aerial video for mapping and 3D urban modeling. A secondary benefit of the proposed work is the open source toolkit that will be released containing all of the developed technology. The existence of this toolkit will benefit the research community, and the research community will, in turn, enhance the software at no cost to the program. The open source toolkit will also directly benefit Kitware by creating more consulting contracts surrounding the technology.