Automatic ship detection in SAR images

The automatic interpretation of Synthetic Aperture Radar (SAR) images is one of the most interesting application field of image processing. This paper addresses the problem of detecting ships in SAR images in a fully automated way. In order to obtain a high reliability and robustness, the proposed processing chain detects possible targets by first searching in parallel for bright spots, i.e. potential ship bodies, and elongated wakes that are present in the scene, and then by cross-validating the wakes against the bright spots in order to reject false targets. This system produces fuzzy decisions, as it detects bright spots and wakes and then associates to each entity a degree of goodness which is determined on the basis of suitable fuzzification functions that are set up during the system training. The whole chain has been implemented and tested on a MicroVax II sequential machine and on a hypercube architecture of IMS T800 transputers.