Scenario Based Transformations for Compensation of Non-Stationary Radar Clutter

Preprint, 2022

Space-Time Adaptive Processing is an important technique for enhancing detection performance in airborne radar systems. The enhanced performance is obtained by mitigating the influence of interference and noise in the radar observations. To perform the mitigation, an accurate estimate of the corresponding space-time covariance matrix of the interference and noise distribution is required. Usually, such an estimate is obtained from secondary data collected from neighboring range bins around the currently investigated cell-under-test. However, in a bistatic radar configuration, the secondary data suffers from geometry-induced angle-Doppler variations along the range dimension. In such configurations, additional processing to handle the angle-Doppler variations is required to obtain a covariance matrix estimate of high accuracy. In this paper, we derive a transformation matrix framework to compensate for the variations over range in the secondary data. The framework is a combination of an incomplete scenario model and secondary data which are used together to obtain a space-time covariance matrix estimate. Thus, the incomplete scenario model is used to find the unitary transformation matrix which, in a Frobenius norm sense, minimizes the expected clutter response from the incomplete scenario model in each range bin towards the corresponding clutter response in a reference range bin. The unitary property of the transformation preserve the stationary behavior of the thermal noise under the transformation. Using such transformation, a set of non-stationary secondary data can be transformed to become more stationary distributed after the transformation. A sample covariance matrix estimator is applied on the transformed set of secondary data to obtain a space-time covariance matrix estimate. The outlined procedure is denoted as a Scenario Based Transformation (SBT) STAP. In numerical simulations, the SBT algorithm is compared with other state-of-the-art methods for the considered problem. The numerical simulations include evaluations on scenarios with a various degree of mismatch between the model generating observations and the model assumed by the investigated algorithms. The included model misspecifications are intrinsic clutter motion, antenna array calibration residuals and incorrect antenna gain patterns. In case of a model match, the simulations indicated that the SBT method yields an improved performance compared to the other investigated methods. For the simulations including model misspecifications, the results indicates that the level of misspecification influence the performance of the considered methods. For a low level of misspecifications, the SBT approach yields an accurate covariance estimate. However, for large misspecifications, the simulations indicates that a non-parametric approach leads to better results.