Acquisition and mosaicing of low-contrast background radiometric images for scene simulation

High resolution radiometric images of a large terrain were obtained by scanning the area with a downlooking imaging radiometer attached to a helicopter. A detailed description of the data acquisition process is presented. The output of this process was thousands of small images, which were then combined (mosaiced) to one large radiometric map. For the mosaicing process we use image registration algorithms developed especially for the case where the input images contain no significant features or contrasts, as might be expected from radiometric desert backgrounds at night. In this situation, traditional feature-based or intensity-based registration methods can give unsatisfactory results. We improve these methods using some heuristics. This enables the automatic mosaicing of a few thousand images in a very short time. An example of the use of the resulting radiometric map for the simulation of a thermal sensor searching for targets is described.