Analysis of a rank-based radar detection system operating on real data

The real versus theoretical behavior of the Federal Aviation Administration's new rank-based digital radar detector, the SRAP I, used in conjunction with their airport surveillance radars, is examined. By examining the behavior of the amplitude ranking process applied to the radar video as it is received in range, it is shown that the expected distribution-free type properties are achieved for each of the Airport Surveillance Radar video types. It is shown that the SRAP I's azimuth sequential observer 'integration' process operating in azimuth on the threshold rank outputs has a random component that can reasonably be modeled as a first-order Markov process which depends on the 'single lag' azimuth correlation of the threshold rank outputs. It is also shown, however, that there is another non-Markov component of false alarms, due in part to surface clutter and vehicular traffic in some regions, for which it is desirable to complement the correlation-based SRAP I control process with direct recursive control. Using each of the three Airport Surveillance Radar video types operating on regions containing some clutter, it is shown that the SRAP I's rank-based detection process outperforms the more standard mean level amplitude comparison-based detection process in terms of standard performance by about 3 dB signal to noise while being controlled to a probability of false alarm level of about 0.00001.