Nonparametric background model based clutter map for X-band marine radar

In a radar system, clutter means any echoes which are not scattered by the wanted target. Usually, the radar clutter map stores an average level for each point or cell in the range-azimuth coordinates as the reference value. A target is then detected in a range-azimuth region if the new echo value in that region exceeds the average background level. In visual computing domain, background model is employed for foreground segmentation, motion detection, salient feature detection, etc. A few kinds of background models are built on each pixel of the sequential images, including the recently popular non-parametric model. In this paper, we proposes a non-parametric background model to implement the radar clutter map. A set of intensity values, which are selected in the past in each pixel location, are stored as the initial clutter model. Then, the new value of a pixel is classified as foreground stemming from a moving target, if the value was stronger than those of the reference samples in the recorded set Clutter model updating is based on randomly choosing in temporal and substituting background pixel values in spatial. Proposed model is proved to be efficient and effective in a moving vehicle detecting application. Also it is compared to a broadly used clutter map method which employs averaging in temporal.