Kitware to Develop Advanced Sparse Bundle Adjustment System for the Air Force

Share Article

Kitware’s sparse bundle adjustment platform will enable more efficient and accurate video analysis by improving camera pose estimation from imagery using advanced algorithms and extensive computer vision expertise.

News Image
Kitware will develop innovative software for performing sparse bundle adjustment for full motion video (FMV) and wide-area motion imagery (WAMI).

Kitware, a leader in developing advanced computer vision technologies, today announces new Phase I SBIR funding from the U.S. Air Force Research Laboratory to develop innovative software for performing sparse bundle adjustment for full motion video (FMV) and wide-area motion imagery (WAMI).

Sparse bundle adjustment (SBA) is a technique for refining 3D scene geometry and the parameters of relative camera motion using images taken from different viewpoints. This technique has existed since the 1950s, but it is only in the past decade that the computational resources required to perform large-scale spare bundle adjustment have become available and practical to use.

Existing open-source toolkits for SBA are slow and focus on large collections of unordered images. Significant optimizations can be made for SBA by exploiting properties found in aerial video, particularly temporal continuity. These optimizations are available at all stages of the camera calibration pipeline, from feature detection and matching to SBA itself.

With this new Phase I funding, Kitware will develop an SBA system targeted for aerial video that will produce state-of-the-art camera calibration accuracy in a fraction of the time compared to current unordered image collection software. A variation of the techniques developed under this project can also be used to construct a streaming SBA solution for live video. Furthermore, relaxation of the specific constraint that bundle adjustment accuracy be optimal over all images could enable real-time, streaming video feeds onboard military vehicles for intelligence, surveillance, and reconnaissance (ISR) systems.

To achieve these goals, Kitware will leverage its world-renowned computer vision expertise, algorithms, and open-source libraries to perform feature detection, feature matching, and sparse optimization.

“We’re excited for the opportunity to improve a well-known yet under-utilized functionality and make a significant impact on video analyst workflows,” said Matthew Leotta, R&D Engineer at Kitware and Principal Investigator on this project. “An additional benefit is that our newly developed system will build upon open-source software and our developed SBA toolkit will also be released to the public as open source.”

To learn more about Kitware’s computer vision expertise and how it can be leveraged to benefit your research, please visit our website.

This work is supported by the Air Force Research Laboratory under Contract No. FA8650-13-M-1694. This contract was approved on September 5, 2013 with FY 13 funds. The views, opinions and/or findings contained in this report are those of the author(s) and should not be construed as an official AFRL position, policy, or decision unless so designated by other documentation.

About Kitware

Kitware is an advanced technology, research, and open-source solutions provider for research facilities, government institutions, and corporations worldwide. Founded in 1998, Kitware specializes in research and development in the areas of visualization, medical imaging, computer vision, quality software process, data management, and informatics. Kitware is headquartered in Clifton Park, NY, with offices in Carrboro, NC, Santa Fe, NM, and Lyon, France. More information can be found at

Share article on social media or email:

View article via:

Pdf Print

Contact Author

Katie Osterdahl
since: 09/2008
Follow >
Follow us on
Visit website