Background clutter suppression and dim moving point targets detection using nonparametric method

A nonparametric background clutter elimination method as an important counterpart of target detection procedures is presented. In our method, after the rejection of clutter, the original spatially correlated imaging sensor data are reduced to independent identical distribution random variables from pixel to pixel that the Gaussianity of residuals is not required while various detection methods strongly rely on it. In this case where the statistics of the residuals (targets or noise) are not known in sufficient details, the employment of a nonparametric detector is often desirable. For the continuity of problem solving, in the second part of this paper, we develop a kind of nonparametric detection algorithm based on track-before-detect method with the capability of detecting targets moving at various speeds. According to the central limit theorem, we derive a likelihood ratio decision rule on the basis of Neyman-Pearson criteria. Computer simulations based on infrared image data with artificially implanted point targets are used for checking the theoretical results.