Improved AR Model-Based Detectors for Range-Spread Targets in Scenarios with a Small Number of Pulses

High-resolution radar systems encounter a challenge where dispersed backscatter energy from targets results in a non-uniform power distribution across range cells. Although autoregressive (AR) modeling of heterogeneous clutter can enhance the probability of range-spread target detection, the performance of detectors significantly decreases in the case of a small number of pulses due to the discard of samples equivalent to the AR order. Therefore, to mitigate performance degradation, this paper presents improved AR model-based detectors for range-spread targets in heterogeneous clutter environments, derived under the assumption of known clutter covariance matrix based on the Rao test, Wald test and generalized likelihood ratio test. Furthermore, the covariance matrix is reconstructed using estimated AR parameters through the relationship between the AR process and triangular matrix decomposition. Additionally, the asymptotic expressions for the probability of detection and false alarm show the new detectors are asymptotically constant false alarm rate with respect to the clutter covariance matrix. Experiments are conducted on both simulated and real clutter data to validate the performance of the newly derived AR model-based detectors. Both sets of results demonstrate that the enhanced detectors maintain robust performance even with a limited number of pulses, outperforming the conventional AR model-based detectors in such scenarios.