Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multi-Domain Features Guided Supervised Contrastive Learning for Radar Target Detection

Junjie Wang, Yuze Gao, Dongying Li, Wenxian Yu

TL;DR

This work tackles radar target detection in dynamic sea clutter by integrating six multi-domain shallow features with deep representations through a supervised contrastive learning framework (MDFG_SCL). The method employs Gini-based weighting of shallow features, a 1D-ResNet50 encoder with a two-layer projection head, and dual losses (matching and supervised contrastive) to align shallow and deep modalities during pre-training, followed by fine-tuning. Experiments on the IPIX1993 dataset show MDFG_SCL achieves superior detection performance and lower false alarm rates, with clear ablation evidence that weighted shallow features and balanced alignment improve results, and strong generalization across diverse datasets and polarization channels. The approach offers practical benefits for robust maritime surveillance, highlighting the value of knowledge-driven feature fusion in challenging radar environments.

Abstract

Detecting small targets in sea clutter is challenging due to dynamic maritime conditions. Existing solutions either model sea clutter for detection or extract target features based on clutter-target echo differences, including statistical and deep features. While more common, the latter often excels in controlled scenarios but struggles with robust detection and generalization in diverse environments, limiting practical use. In this letter, we propose a multi-domain features guided supervised contrastive learning (MDFG_SCL) method, which integrates statistical features derived from multi-domain differences with deep features obtained through supervised contrastive learning, thereby capturing both low-level domain-specific variations and high-level semantic information. This comprehensive feature integration enables the model to effectively distinguish between small targets and sea clutter, even under challenging conditions. Experiments conducted on real-world datasets demonstrate that the proposed shallow-to-deep detector not only achieves effective identification of small maritime targets but also maintains superior detection performance across varying sea conditions, outperforming the mainstream unsupervised contrastive learning and supervised contrastive learning methods.

Multi-Domain Features Guided Supervised Contrastive Learning for Radar Target Detection

TL;DR

This work tackles radar target detection in dynamic sea clutter by integrating six multi-domain shallow features with deep representations through a supervised contrastive learning framework (MDFG_SCL). The method employs Gini-based weighting of shallow features, a 1D-ResNet50 encoder with a two-layer projection head, and dual losses (matching and supervised contrastive) to align shallow and deep modalities during pre-training, followed by fine-tuning. Experiments on the IPIX1993 dataset show MDFG_SCL achieves superior detection performance and lower false alarm rates, with clear ablation evidence that weighted shallow features and balanced alignment improve results, and strong generalization across diverse datasets and polarization channels. The approach offers practical benefits for robust maritime surveillance, highlighting the value of knowledge-driven feature fusion in challenging radar environments.

Abstract

Detecting small targets in sea clutter is challenging due to dynamic maritime conditions. Existing solutions either model sea clutter for detection or extract target features based on clutter-target echo differences, including statistical and deep features. While more common, the latter often excels in controlled scenarios but struggles with robust detection and generalization in diverse environments, limiting practical use. In this letter, we propose a multi-domain features guided supervised contrastive learning (MDFG_SCL) method, which integrates statistical features derived from multi-domain differences with deep features obtained through supervised contrastive learning, thereby capturing both low-level domain-specific variations and high-level semantic information. This comprehensive feature integration enables the model to effectively distinguish between small targets and sea clutter, even under challenging conditions. Experiments conducted on real-world datasets demonstrate that the proposed shallow-to-deep detector not only achieves effective identification of small maritime targets but also maintains superior detection performance across varying sea conditions, outperforming the mainstream unsupervised contrastive learning and supervised contrastive learning methods.

Paper Structure

This paper contains 15 sections, 6 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Multi-domain features extraction
  4. Supervised contrastive learning method for radar target detection
  5. Data Augmetation
  6. Deep Feature Extraction
  7. Match Loss Function and Supervised Contrastive Loss Function
  8. Experiments and Discussion
  9. Datasets Description
  10. Experiment Setting
  11. Results and Discussion
  12. Performance Comprison
  13. Ablation Experiments
  14. Generalization
  15. Conclusion

Figures (4)

  • Figure 1: Supervised contrast learning framework based on multi-domain shallow features guidance.
  • Figure 2: Detection performance mIoU at $P_{fa}$=0.001.
  • Figure 3: Evaluation metrics change under different $\alpha$.
  • Figure 4: Detection visualization results at $P_{fa}$=0.01. (a) (d)Detection results of traditional supervised learning, unsupervised contrastive learning, supervised contrastive learning, and the proposed method on dataset #280. (e) (h)Detection results on dataset #310. (i) (l)Detection results on dataset #311. (m) (p) Detection results on dataset #320.