Rejection of road clutter using mean-variance method with OS-CFAR for automotive applications

In order to improve the accuracy of target detection, two main conditions are required. One is the high signal to interference ratio (SIR) and the other is the signal processing that classifies signals into echo from targets and clutters accurately. Constant false alarm rate (CFAR) is the step to classify them by adaptive threshold value. However, there are limitations because of the unexpected power of clutter signals which cause a target missing and a false alarm. In this paper, we propose additional simple step after order statistic-CFAR (OS-CFAR) to reduce the false alarm. It can be analyzed by the mean-variance of the neighboring cells of the target declared. There are two major assumptions. First, clutters spread its power wider than targets´ power. Next, reflected power from targets is larger than from clutters. These assumptions are supported by real data sets obtained with 24GHz near-range radar designed for frontal collision mitigation. Experimental result shows that false alarms are reduced by new scheme.