Predicting False Alarm Rates for High-Resolution Antisubmarine Warfare Sonars in a Cluttering Environment Prone to False Alarm Rate Inflation

Abstract

Operation of high-resolution, broadband, antisubmarine warfare sonars in littoral waters is challenging, since the presence of sea mounts, underwater ridges, and other topographic features causes increased false alarm rates. Two important contributors to the raised false alarm rate are the signal-processing induced phenomenon called false alarm rate inflation (FARI) and the presence of sonar clutter, also referred to as non-Rayleigh distributed matched filter envelope in literature. Conventional constant false alarm rate (CFAR) algorithms fail to achieve a constant false alarm rate in all ranges and bearing in the presence of such effects. Given sufficient information on the bathymetry and the bottom properties, the occurrence of FARI may be estimated through the use of an acoustic model. This allows for more accurate estimates of the false alarm rate. Through measurements, the scatterer statistics for a given sonar in a given area may be estimated. Combining a FARI predicting scheme with knowledge on the scatterer statistics allows the estimation of the probability of false alarm in the presence of both FARI and sonar clutter. Here, we propose a new detection scheme that employs this approach to estimate a range- and bearing-dependent threshold that can be applied on normalized sonar data to achieve a CFAR even in the presence of FARI and clutter. The performance of the method is assessed through the use of receiver operating characteristic curves and is shown to outperform conventional CFAR algorithms, such as the cell averaging, greater of, and ordered statistics CFAR processors. The method is tested on both recorded and synthetic data. The robustness of the method is tested using synthetic data by introducing errors in the topography, sound speed, and scatterer statistics when estimating the probability of false alarm. The performance of the method decreases when introducing these errors, but it still outperforms the conventional CFAR processors.