Characterization of a Bayesian Ship Detection Method in Optical Satellite Images

This letter presents the experimental results obtained for an automatic predetection of small ships (about 5 $\times$ 5 pixels) in high-resolution optical satellite images. Our images are panchromatic SPOT 5 images, whose resolution is 5 m per pixel. Our detection method is based on the Bayesian decision theory and does not need any preprocessing. Here, we describe the method precisely and the tuning of its two parameters, namely, the size of the analysis window and the threshold used to make a decision. Both are fixed from the receiver operating-characteristic curves that we draw from different sets of tests. Finally, the overall results of the method are given for a set of images, as close as possible to the operational conditions.