Remote sensing of moving targets, particularly pedestrians on the road, is crucial for advanced driver assistance systems. However, pedestrian detection using the radar system remains an ongoing challenge due to the radar cross section (RCS) of pedestrians being much smaller than that of the clutter. Existing radar systems and pedestrian detection methods predominantly rely on the single-polarization radar, while research on the fully polarized radar for pedestrian detection is relatively limited. In this paper, the L-band fully polarimetric radar system is developed for pedestrian detection, and based on the full polarized radar echo HH, HV, VH, and VV, a novel clutter suppression method is proposed, which integrates the optimal polarization states of antennas and optimal scattering characteristics of pedestrians. Moreover, the field experiment has been conducted, and the results demonstrate that the signal-to-clutter-plus-noise ratio (SCNR) of the total power signal of full-polarization echoes is higher than that of single-polarization echoes, and the proposed clutter suppression method is able to reduce the non-stationary clutter and the interference signal generated by the multipath effect, thereby improving the SCNR. Furthermore, the OTSU algorithm is employed to detect pedestrian targets using radar data before and after clutter suppression, and the results demonstrate that the proposed method yields superior detection performance. These findings justify the potential of fully polarimetric radar in enhancing pedestrian detection.

The interference of sidelobe often causes different targets to exhibit similar features, diminishing fine-grained classification accuracy. This effect is particularly pronounced when the available data are limited. To address the aforementioned issues, a novel classification framework for sidelobe-affected SAR imagery is proposed. First, a method based on maximum median filtering is adopted to remove sidelobe by exploiting local grayscale differences between the target and sidelobe, constructing a high-quality SAR dataset. Second, a deep metric learning network is constructed for fine-grained classification. To enhance the classification performance of the network on limited samples, a feature extraction module integrating a lightweight attention mechanism is designed to extract discriminative features. Then, a hybrid loss function is proposed to strengthen intra-class correlation and inter-class separability. Experimental results based on the FUSAR-Ship dataset demonstrate that the method exhibits excellent sidelobe suppression performance. Furthermore, the proposed framework achieves an accuracy of 84.18% across five ship target classification categories, outperforming the existing methods, significantly enhancing the classification performance in the context of sidelobe interference and limited datasets.

The Global Navigation Satellite System-Scattering (GNSS-S) radar has a wide coverage and strong concealment, enabling large-scale and long-term monitoring of sea surface targets. However, its signal power is extremely low and susceptible to sea clutter interference. To address the challenge of detecting and tracking moving targets in complex maritime environments using low-resolution radar, this paper proposes a method for extracting moving target trajectories from GNSS-S radar under low signal-to-noise ratio (SNR) conditions. The method constructs a feature plane consisting of statistical and polarization characteristics, based on the unique distribution of different motion targets in this plane, the distinction between sea clutter and multi-motion targets is carried out using machine learning algorithms, and finally the trajectory association of the targets is achieved by the Kalman filter, and the tracking correctness can reach more than 93.89%. Compared with the tracking method based on high-resolution imaging targets, this technique does not require complex imaging operations, and only requires certain processing on the radar echo, which has the advantages of easy operation and high reliability.

Adversarial examples pose a significant threat to Deep Neural Networks (DNNs) underpinning Synthetic Aperture Radar (SAR) Automatic Target Recognition (ATR) systems, as these models exhibit acute susceptibility to such malicious inputs. While white-box attacks achieve high success rates, their transferability to unknown black-box models—particularly across different network architectures (e.g., from CNNs to Vision Transformers)—remains a significant challenge. Existing gradient-based iterative methods often overfit the specific decision boundary of the surrogate model, resulting in poor generalization. To address this, we propose a novel generative attack framework termed BUM. Instead of merely maximizing the classification error, BUM explicitly models and minimizes the epistemic uncertainty of the surrogate model. By leveraging Monte Carlo (MC) Dropout to simulate a Bayesian ensemble, we train a generator to craft perturbations that are consistently adversarial across stochastic sub-models. This regularization forces the attack to target high-level, structure-aware semantic features shared among architectures, rather than low-level, model-specific artifacts. Extensive experiments on the MSTAR and FUSAR datasets demonstrate the superior black-box transferability of BUM.