Signal feature extraction for rain clutter mitigation in surveillance radar

Rain clutter can significantly degrade the performance of surveillance radars by generating false detections and masking returns from genuine targets. In single-polarization architectures, this challenge is amplified by the absence of polarimetric signatures, and existing single-attribute approaches may not reliably separate target and clutter returns. This paper investigates an offline approach for discriminating target and rain-clutter echoes using real-world single-polarization measurements acquired with the SABER M200 VIGILANTE S-band surveillance radar. The proposed classifier combines complementary amplitude-, coherence-, and spectral-domain attributes derived from the Peak-to-Average Power Ratio (PAPR), pair-wise inter-receiver coherence, and background-region Spectral Flatness Measure (SFM) statistics. The dataset comprises measurements collected under different rainy-weather conditions and acquisition scenarios, including situations in which target and rain-clutter returns coexist. To evaluate generalization, entire measurement days are reserved exclusively for testing on a per-class basis, ensuring that training and test data never share the same day and class. Under this evaluation, the method achieves average recall values of 0.94 for target returns and 0.97 for rain clutter echoes. These results support the use of complementary signal-derived attributes for offline rain clutter discrimination in single-polarization surveillance radar measurements.