Autonomous Distant Visual Surveillance of Satellites (PREPRINT)
Abstract
This paper develops three new, interconnected techniques useful for the autonomous distant visual inspection of satellites. First, silhouetting of man made, erratically illuminated satellites is performed. Illumination cases include full sun from an arbitrary (often awkward) viewing angle and unilluminated (back-lit by the star field). New Statistical Straight Line Snakes (SSLS) prove efficient in finding the silhouette, even in the unilluminated case. The silhouette is composed of straight line segments, which are easy to calculate, fit the straight lines inherent in man made objects, and lend themselves to further processing (pose estimation, template matching, etc.). Once the silhouette has been used to find correspondence points, a second method for detecting a moving, nearby chaser vehicle is derived. The hard case is treated in which the chaser and satellite are so nearby that their images are blurred together. The algorithm finds the dimension of motion generated by the sequence of images. If the dimension is higher than that explained by a single rigid body, then this indicates a possible chaser. Independent relative motion between the satellite and chaser is required -- if the chaser is immobile with respect to the satellite, then a third technique must be used. This third method incorporates the satellite's solid model to estimate its pose from a noisy, diffraction limited image. The pose is then combined with the solid and optical model to create synthetic expected images. Inspection is performed by comparing these with the actual images. The new pose algorithm first estimates depth by a least upper bound technique. A fast method is derived of optimally estimating the rotation matrix by a sequence of analytical solutions (rather than a nonlinear numerical optimization!). Simulations illustrate the use of all three techniques on images obtained when viewing low Earth orbit satellites from the ground.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 10, 2006
- Accession Number
- ADA462037
Entities
People
- John E. Mcinroy
- Lawrence M. Robertson
- R. S. Erwin
Organizations
- Air Force Research Laboratory