Airborne multi-spectral images registration through genetic algorithm

The proposed paper deals with images registration through airborne images mosaicking, in the context of environmental monitoring. The matching problem consists, here, in determining the transfer parameters required to partially map one multi spectral image to the adjacent images of the flight lines. The purpose is to build a seamless representation of the regions flown over by the aircraft. However, given the unsteadiness of the airborne platform, a single affine transformation cannot cope with the acquisition distortions and cannot yield to a seamless mosaic. Thus, the principle of the developed approach relies on a field of transformations rather than on a single transformation. Genetic agents are introduced to carry out the local matching transformations. The approach, even if it involves a sequential programming, proved to be efficient from the computational time point of view.