A machine vision algorithm was developed which permits guidance control to be maintained during autonomous proximity operations. At present this algorithm exists as a simulation, running upon an 80386 based personal computer, using a ModelMATE CAD package to render the target vehicle. However, the algorithm is sufficiently simple, so that following off-line training on a known target vehicle, it should run in real time with existing vision hardware. The basis of the algorithm is a sequence of single camera images of the target vehicle, upon which radial transforms were performed. Selected points of the resulting radial signatures are fed through a decision tree, to determine whether the signature matches that of the known reference signatures for a particular view of the target. Based upon recognized scenes, the position of the maneuvering vehicle with respect to the target vehicles can be calculated, and adjustments made in the former's trajectory. In addition, the pose and spin rates of the target satellite can be estimated using this method.
Autonomous Proximity Operations Using Machine Vision for Trajectory Control and Pose Estimation
1991
7 pages
Report
No indication
English
Computer Aided Design, CAD , Pattern Recognition & Image Processing , Algorithms , Autonomy , Computer aided design , Computer vision , Computerized simulation , Decision theory , Maneuvers , Signature analysis , Trajectory control , Cameras , Personal computers , Real time operation , Signatures , Transformations (Mathematics)
Vision-based pose estimation for autonomous operations in aquacultural fish farms
BASE | 2021
|