Issues in advanced image registration for multi-UAV platform applications

Image registration is in some senses the 'forgotten problem' in multi-sensor image exploitation. Image registration will, at present, typically involve using a constant registration transform to align images from sensors fixed relative to each other and separated by as small a distance as possible. For more challenging situations in which sensors are more widely spaced and not fixed relative to each other (e.g. teams of Uninhabited Air Vehicles, UAVs) the registration problem becomes far more complex. The theory of solving the registration problem for such situations is poorly understood with outstanding issues being the expected optimal image alignment, how to perform automatic registration, and methods of addressing time varying image registration. As a pre-cursor to the solution of such problems we present an analysis of the likely errors associated with more complex registration problems. Results of trials using current state-of-the-art technology are presented followed by initial concepts in improving these results.