A CFAR Circuit of Detecting Spatially Correlated Target for Automotive UWB Radars

Automotive radars for potential accident reduction and for autonomous cruse control were available in 1999, and 77 GHz long-range radars and 24 GHz short-range radars are developing [1]. 79-GHz shortrange automotive radars have a bandwidth of 4 GHz, a maximum range of 30 m, and a range resolution of 5 cm. Radar reception signals often contain unwanted echoes, called “radar clutter.” Moving target indication (MTI) is a possible way for detecting a target with a motionless radar and the clutter objects are motionless, too. Automatic target detection with a fixed threshold encounters an increase in a probability that the radar faulty detect a target when there is no target. Constant false alarm rate (CFAR) circuits measure local mean amplitude to suppress an increase in the false alarm rate. CFAR adjusts the threshold adaptively using the average samples [2]. Radar echoes tend to have high peak-to-average power ratio an increase in a resolution, then the the echoes are referred to as spiky. For sea clutter, there are many echoes from sea surface facets within the radar illumination area and the resulting signal obeys Rayleigh distribution because of the central limit theorem. With high resolution radars, however, a few facets contribute the resulting signal and their mutual interferences characterize spiky clutter [3]. The amplitude fluctuation has been modeled as Weibull distribution, log-normal distribution, and K-distribution. Various modifications of CFAR circuits, such as ordered-statistics CFAR, cell-average CFAR, and and-or CFAR, are examined for automotive radars [4]. In this paper, a CFAR circuit that measures local mean powers of both target and clutter is proposed. High resolution automotive radars would receive spiky signals not only for clutter but also for target. Spatial correlation is introduced to evaluate performance of the proposed method, because automotive radars have higher range resolution that target and clutter would occupy several range cells, and spatial correlation increases a required target signal power for detection.