Clutter characterization for robust detection of slow moving targets in Ka-band Noise Radar Images

In this paper, real data acquired by the Ka-band Ground Based Noise Waveform SAR sensor designed and developed at the Laboratory for Nonlinear Dynamics of Electronic Systems of the Usikov Institute for Radiophysics and Electronics of NASU, is analyzed to propose statistical clutter models for the design of improved robust detection schemes. This high resolution surface radar is expected to discover the non-homogeneity of ground clutter, requiring a detection solution with adaptation capabilities against different clutter sources and time variability caused by weather conditions, among others. After a detailed statistical analysis for intensity modelling, Gamma and Generalized Gamma distributions (GΓD) were proposed for ground clutter (for example, grass or asphalt), and with different types of trees, respectively. Constant False Alarm Rate (CFAR) detectors were designed assuming these distributions and compared to a conventional CFAR for Gaussian clutter. Results prove a better performance of the proposed detectors according to Probability of False Alarm requirements, compared to the Gaussian one. The GΓD-CFAR provided the best results: the probability of false alarm was the closest to the desired one, with a slight reduction in probability of detection of slow moving targets.