A Multi-Dimensional Feature Enhancement Network for SAR Target Detection via Cascaded Frequency–Spatial Refinement.
Saved in:
| Title: | A Multi-Dimensional Feature Enhancement Network for SAR Target Detection via Cascaded Frequency–Spatial Refinement. |
|---|---|
| Authors: | Guo, Shanhong1 (AUTHOR) guosh@njust.edu.cn, Zhu, Ji1,2 (AUTHOR), Chen, Gao1,2 (AUTHOR), Yang, Mu1,2 (AUTHOR), Sheng, Weixing1 (AUTHOR) |
| Source: | Remote Sensing. Jun2026, Vol. 18 Issue 12, p1888. 32p. |
| Subjects: | Wavelet transforms, Feature extraction, Radar targets, Speckle interference, Deep learning, Military reconnaissance |
| Abstract: | Highlights: What are the main findings? A frequency–spatial decoupling block (FSDBlock) is proposed that cascades wavelet convolution with dynamic spatial-adaptive convolution, effectively separating speckle noise from target edge features in the frequency domain while adapting to geometric non-stationarity in the spatial domain. The proposed FSDNet achieves state-of-the-art performance on both HRSID (mAP50 = 92.3%, mAP50:95 = 68.6%) and SSDD (mAP50 = 98.7%, mAP50:95 = 74.2%) SAR ship detection benchmarks, consistently outperforming existing methods. What are the implications of the main finding? The frequency–spatial decoupling strategy provides a physically interpretable paradigm for SAR image feature extraction, demonstrating that explicit frequency-domain decomposition can resolve the long-standing conflict between noise suppression and texture preservation in SAR target detection. The lightweight architecture (6.21M parameters) and consistent improvements across datasets suggest strong potential for practical deployment in real-time SAR reconnaissance and maritime surveillance applications. Target detection in synthetic aperture radar (SAR) images is constrained by three primary challenges. First, speckle noise overlaps heavily with the high-frequency features of target edges in the frequency domain, so standard convolutions cannot suppress noise without sacrificing edge texture. Second, the scattering signature of a SAR target varies markedly with viewing angle, and a fixed-parameter convolution kernel cannot accommodate this spatial non-stationarity. Third, deep and shallow levels of the feature pyramid differ in semantics and resolution, and a naive element-wise sum either introduces noise interference or loses small-target signals. We propose the Frequency–Spatial Detection Network (FSDNet), whose core FSDBlock cascades three operators to address these failure modes in turn. Wavelet Convolution (WTConv) projects features into Haar sub-bands and applies independent low- and high-frequency kernels prior to inverse-DWT reconstruction, suppressing noise while preserving edges. Receptive-Field Attention Convolution (RFAConv) generates location-conditional kernels and so adapts to non-stationary scattering. Spatial Context Self-Attention (SCSA) aggregates discrete scattering points into coherent target representations via long-range grouped attention. At the fusion stage, CGAFusion replaces FPN element-wise addition with a channel–spatial–pixel triple-attention soft switch that mitigates deep–shallow semantic mismatch. On HRSID, FSDNet attains mAP50 = 92.3 % and mAP50:95 = 68.6 %. On SSDD, it attains mAP50 = 98.7 % and mAP50:95 = 74.2 %. Both sets of results consistently surpass the baseline methods. Against the strongest YOLO baseline (YOLOv11n), FSDNet improves HRSID mAP50 by + 1.7 percentage points (pp) and mAP50:95 by + 2.3 pp, and SSDD mAP50 by + 0.5 pp and mAP50:95 by + 2.7 pp; against the capacity-fair YOLOv11s reference (∼51% more parameters), FSDNet still leads on mAP50, mAP50:95, recall, and F1. Ablation studies and power-spectral-density analyses corroborate the contribution of each module and confirm WTConv's role in preserving high-frequency target features. [ABSTRACT FROM AUTHOR] |
| Copyright of Remote Sensing is the property of MDPI and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| Abstract: | Highlights: What are the main findings? A frequency–spatial decoupling block (FSDBlock) is proposed that cascades wavelet convolution with dynamic spatial-adaptive convolution, effectively separating speckle noise from target edge features in the frequency domain while adapting to geometric non-stationarity in the spatial domain. The proposed FSDNet achieves state-of-the-art performance on both HRSID (mAP50 = 92.3%, mAP50:95 = 68.6%) and SSDD (mAP50 = 98.7%, mAP50:95 = 74.2%) SAR ship detection benchmarks, consistently outperforming existing methods. What are the implications of the main finding? The frequency–spatial decoupling strategy provides a physically interpretable paradigm for SAR image feature extraction, demonstrating that explicit frequency-domain decomposition can resolve the long-standing conflict between noise suppression and texture preservation in SAR target detection. The lightweight architecture (6.21M parameters) and consistent improvements across datasets suggest strong potential for practical deployment in real-time SAR reconnaissance and maritime surveillance applications. Target detection in synthetic aperture radar (SAR) images is constrained by three primary challenges. First, speckle noise overlaps heavily with the high-frequency features of target edges in the frequency domain, so standard convolutions cannot suppress noise without sacrificing edge texture. Second, the scattering signature of a SAR target varies markedly with viewing angle, and a fixed-parameter convolution kernel cannot accommodate this spatial non-stationarity. Third, deep and shallow levels of the feature pyramid differ in semantics and resolution, and a naive element-wise sum either introduces noise interference or loses small-target signals. We propose the Frequency–Spatial Detection Network (FSDNet), whose core FSDBlock cascades three operators to address these failure modes in turn. Wavelet Convolution (WTConv) projects features into Haar sub-bands and applies independent low- and high-frequency kernels prior to inverse-DWT reconstruction, suppressing noise while preserving edges. Receptive-Field Attention Convolution (RFAConv) generates location-conditional kernels and so adapts to non-stationary scattering. Spatial Context Self-Attention (SCSA) aggregates discrete scattering points into coherent target representations via long-range grouped attention. At the fusion stage, CGAFusion replaces FPN element-wise addition with a channel–spatial–pixel triple-attention soft switch that mitigates deep–shallow semantic mismatch. On HRSID, FSDNet attains mAP50 = 92.3 % and mAP50:95 = 68.6 %. On SSDD, it attains mAP50 = 98.7 % and mAP50:95 = 74.2 %. Both sets of results consistently surpass the baseline methods. Against the strongest YOLO baseline (YOLOv11n), FSDNet improves HRSID mAP50 by + 1.7 percentage points (pp) and mAP50:95 by + 2.3 pp, and SSDD mAP50 by + 0.5 pp and mAP50:95 by + 2.7 pp; against the capacity-fair YOLOv11s reference (∼51% more parameters), FSDNet still leads on mAP50, mAP50:95, recall, and F1. Ablation studies and power-spectral-density analyses corroborate the contribution of each module and confirm WTConv's role in preserving high-frequency target features. [ABSTRACT FROM AUTHOR] |
|---|---|
| ISSN: | 20724292 |
| DOI: | 10.3390/rs18121888 |
