Hongtao Su

The detection of small sea-surface targets remains challenging due to their low signal-to-clutter-plus-noise ratios (SCNR), and low Doppler frequencies. Conventional adaptive detectors fail to accurately model sea clutter, resulting in high false alarm rates and reduced detection probability. Feature-based methods avoid explicit clutter modeling but struggle to extract sufficiently discriminative features and generally require long observation durations. Although deep learning (DL) methods offer a promising alternative, most existing DL detectors rely on single-range-cell feature extraction, limiting their ability to distinguish weak targets from complex clutter backgrounds. In practice, sea clutter across adjacent range cells is typically independently and identically distributed (i.i.d.). Consequently, the autocorrelation features of a target-containing cell under test will exhibit distinct discrepancies compared with those of reference cells. Accordingly, we propose a detection method that leverages both the temporal correlation of the echo signal within each range cell and the similarity of their autocorrelation information between the cell under test and reference cells. While standard Transformers are adept at sequence processing, their inherent architecture cannot explicitly model such intra-cell autocorrelation. To realize the proposed mechanism, we introduce a neural network framework that employs a Multi-Dimensional Attention Autocorrelation (MDAttAC) module for explicit feature extraction. Subsequently, a standard Transformer is cascaded to evaluate the spatial similarity of these robust features across range cells. Experimental results demonstrate that the proposed method significantly enhances detection performance, reduces reliance on long observation durations, and exhibits superior generalization capability.

Multimodal remote sensing combines optical and synthetic aperture radar (SAR) imagery to improve perception, yet real deployments face spatially varying degradations (e.g., clouds, low light, sensor interference) that can corrupt fusion. To make robustness measurable, we introduce a controlled mixed-severity setting in which only the optical stream is synthetically cloud-degraded while SAR remains intact, providing a standardized testbed for evaluating multimodal detection under modality imbalance. We further present CAIR-Net, a reliability–aware information routing network that follows a denoise-then-fuse principle: a Local Reliability Modulation (LRM) module learns soft, spatial reliability maps to suppress degraded regions before cross-modal interaction, and a Global Information Selection Mechanism (GISM) performs confidence-aware expert routing across optical, fused, and SAR experts. On the mixed-severity benchmark, CAIR-Net consistently outperforms strong unimodal and fusion baselines and exhibits a substantially smaller performance drop under severe clouds (only a 7.3% AP reduction versus drops exceeding 25% for representative alternatives). These results indicate that explicit reliability modeling and quality-guided routing provide a practical path toward robust multimodal detection when one modality is partially or nearly completely occluded.

This paper considers the application scenarios of dual-function radar communication (DFRC) systems with multiple communication users and multiple detection targets. Generally, communication requires the transmission of data to users, whereas the task of radar is to get the target parameters by analyzing echoes. Therefore, we jointly optimize the radar-receiving filter and dualfunction waveform in the integrated sensing and communication system. Unlike existing research, which introduced a trade-off parameter to integrate radar and communication functions into one objective function, we minimize communication multiuser interference (MUI) energy under radar signal-to-interference-plusnoise ratio (SINR) constraints. Additionally, a low peak-to-averagepower ratio (PAPR) constraint is imposed on the waveform to avoid nonlinear distortion after amplification. According to the alternating direction method of multipliers (ADMM) framework, a cyclic algorithm is presented to seek the solution of the multiconstraint nonconvex optimization problem. In addition, we discuss the proposed method from computational load and convergence. The numerical results demonstrate that the designed filter and waveform enable the DFRC system to achieve satisfactory communication and detection capabilities