Tracking Of Translated And Rotated FeaturesWithin Satellite Images Using A TightlyCoupled MIMD Computer

A technique for establishing sea surface velocities in sequential satellite infrared images has been developed using a multiprocessor computer system. A consecutive pair of Sea Surface Temperature (SST) images is produced from satellite-derived data and brought into register. A Maximum Cross Correlation (MCC) approach is then used to determine the amount of sea surface feature movement over time. The amount of feature movement, and the time interval between images, determine the surface velocity. As the water patterns observed tend to exhibit rotational movement as well as translation, the MCC method has been extended to detect rotational motions. In general, the correlation coefficient is computed for all possible translations and rotations of the MCC template within the search area. Depending on the size of the template in use, a small amount of template rotation may have little or no effect on the correlation coefficient. It is accordingly possible to use an adaptive approach to minimise correlation searches. By utilizing an Alliant FX/2800 shared memory MIMD computer, and applying appropriate algorithmic and compiler optimisation, the task is implemented sufficiently fast that a complete satellite overpass may be processed before the next co-incident satellite orbit. The system implemented suggests that the operational tracking of rotating sea features is possible in a multiprocessor machine environment.