We propose an algorithm for aimpoint maintenance that is robust to large changes in target aspect as well as scintillation and atmospheric turbulence. The approach makes use of flexible, parametrized vehicle models that are capable of representing a vast array of potential targets. Information about the platform orientation collected from on-board sensors and minimal user input provide initial constraints, and the model's structure parameters are refined during the course of the track. As the structure parameters converge, the motion of the target is tracked by comparing the model to the camera output in order to determine the relative position and orientation of the target. During Phase I, we will develop a prototype implementation of the algorithm and demonstrate its efficacy on imagery provided by the government.
Benefit: This technology will be useful for the engagement of ground and maritime targets in military and law enforcement. Suppression and surveillance applications are possible.
