SAR Image Ship Target Detection under Complex Background

As the first step of the automatic image interpretation system, the automatic detection of the target must be accurate and fast. For synthetic aperture radar (SAR) images, CFAR detection algorithm is the most commonly used target detection framework. In CFAR algorithm, the modeling of background clutter is very important, because the detection threshold is calculated based on this model. The Gumbel Mixed (GM) distribution model is proposed to model the background clutter, and the expected maximization method (EM) is used to estimate the parameters of the model, and Newton iteration method is applied to obtain the detection threshold. The model is effective in modeling complex statistics, including but not limited to statistics involving heavy tail distribution. At the same time, os-cfar detection algorithm of ordered statistics based on the model was applied to detect ship targets in SAR images. The experimental data set was derived from public SAR images, and the effect was significantly improved compared with traditional methods.