浏览全部资源
扫码关注微信
收稿：2026-04-16，
修回：2026-07-17，
录用：2026-08-10，
网络首发：2026-08-20，
移动端阅览
目的
2
针对高分辨率极化合成孔径雷达（Polarimetric Synthetic Aperture Radar， PolSAR）图像由于空间非平稳性和散射异质性导致的边界混淆及区域语义不一致问题，提出了一种语义划分与差异化对比学习的 PolSAR 图像分类方法，旨在提升模型对复杂地物的判别能力和边界保持能力。
方法
2
首先，以层次语义模型（Hierarchical Semantic Model， HSM）为理论指导，结合超像素分割技术对图像进行语义划分，精准划分边界敏感区域与非边界稳定区域并构建掩码引导机制；其次，设计双分支异构特征提取网络，利用卷积神经网络（Convolutional Neural Network， CNN）捕获边界区域的细粒度散射特征，同时借助基于超像素的图卷积网络（Superpixel-based Graph Convolutional Network， SGCN）建模非边界区域的长程空间依赖关系；最后，引入自适应边界权重的差异化对比学习策略，通过对边界难分样本的强化监督和跨区域特征对齐，实现特征空间的协同优化。
结果
2
在三组公开极化 SAR 数据集上进行了实验验证。在西安数据集中，本文算法的总体准确度（Overal Accuracy， OA）达到 94.58%，相较于其他主流对比方法，OA提升了0.57%至5.48%；在Flevoland数据集中，OA达到 99.63%，且在绝大多数地物类别上取得了最优分类精度；在Oberpfaffenhofen数据集中，OA达到98.65%，显著缓解了郊区类别的错分问题。参数分析表明，5%的训练样本比例可有效平衡精度与人工标注成本。
结论
2
本文提出的分类方法通过“语义分区-异构建模-对比增强”的深度融合，有效解决了传统方法中区域信息与特征学习脱节的瓶颈，在显著提升分类精度的同时，表现出优良的边界保持能力与空间一致性。
Objective
2
High-resolution Polarimetric Synthetic Aperture Radar （PolSAR） image classification is a critical task in remote sensing， providing essential data for land-use monitoring， disaster assessment， and environmental resource management. PolSAR imagery characterizes the physical scattering mechanisms and spatial geometric structures of ground objects through multi-channel coherent observations. However， high-resolution PolSAR images often exhibit significant spatial non-stationarity and scattering heterogeneity due to the inherent imaging mechanism and the interference of speckle noise. Traditional deep learning classification paradigms predominantly utilize isotropic feature processing， applying unified convolutional kernels or network structures across the entire spatial domain. Such models fail to account for the essential differences between boundary and non-boundary pixels in terms of statistical distribution and spatial correlation. Standard convolutional operators are susceptible to interference from heterogeneous neighboring pixels when processing boundary areas， leading to edge positioning deviations and classification "overflow". Conversely， for large-scale homogeneous non-boundary areas， local feature extraction operators struggle to capture long-range spatial dependencies， limiting the ability to maintain semantic consistency. To address these limitations， this study proposes a novel PolSAR image classification method based on region partitioning and differentiated contrastive learning， aimed at optimizing feature representation for complex heterogeneous scenes.
Method
2
The proposed framework follows a systematic "prior layer-backbone layer-supervision layer" architecture. First， as a topological prior， a Hierarchical Semantic Model （HSM） is employed for spatial deconstruction. Based on Marr's visual computation theory， the HSM interprets structural clues to generate a polarization sketch and subsequently partitions the image into boundary-sensitive regions and non-boundary stable regions （comprising homogeneous and structured heterogeneous areas）. These results are used to generate binary region masks $M（x，y）$ that provide spatial routing for subsequent feature extraction. Second， a dual-branch heterogeneous backbone network is constructed to adapt to the distinct scattering characteristics of these regions. A Convolutional Neural Network （CNN） branch utilizes small receptive fields to capture fine-grained scattering mutations and local details in boundary areas. Simultaneously， a Superpixel-based Graph Convolutional Network （SGCN） is designed for non-boundary areas. By treating superpixels generated by the Pol_ASLIC algorithm as graph nodes and utilizing symmetric revised Wishart distance for edge weighting， the SGCN effectively models long-range spatial dependencies and maintains global semantic consistency. A feature enhancement module precedes the dual branches to suppress speckle noise. The features from both branches are then integrated through a channel concatenation and a Convolutional Block Attention Module （CBAM） to suppress redundant information and enhance discriminative power. Third， an adaptive boundary-importance differentiated contrastive learning mechanism is introduced as the supervision layer. High-dimensional fused features are mapped to a compact contrastive space. The system learns pixel-level adaptive weights， assigning higher importance to boundary pixels to strengthen supervision on difficult-to-classify samples. The total loss function incorporates a cross-entropy classification loss and a differentiated contrastive loss， which applies distinct constraints to boundary， non-boundary， and cross-region samples to optimize the structure of the feature space.
Result
2
The effectiveness of the proposed method was evaluated on three widely used public PolSAR datasets： Xi'an （China）， Flevoland （Netherlands）， and Oberpfaffenhofen （Germany）. We compared our model against five state-of-the-art methods—Super_RF（Superpixel-based Random Forest）， DFGCN（Deep Fuzzy Graph Convolutional Networks）， AMS-M2ESL（Adaptive Mask Sampling and Manifold-Euclidean Subspace Learning）， 3DCNN（Three-Dimensional Convolutional Neural Network）， and PolMPCNN（Polarimetric Multi-Path Convolutional Neural Network）—as well as an ablation model， RPHNet（Region Partition Based Hybrid Deep Network）. In the Xi'an dataset experiment， our method achieved an Overall Accuracy （OA） of 94.58% and a Kappa coefficient of 0.9106， outperforming the second-best method， PolMPCNN， by 0.57% in OA. Quantitative analysis showed significant improvements in grass and building categories. For the Flevoland dataset， which features 15 complex crop types， the proposed method achieved an OA of 99.63% and an Average Accuracy （AA） of 99.57%. It successfully suppressed the salt-and-pepper noise and boundary confusion observed in comparative methods， particularly in the Wheat 2 and Rape categories. In the Oberpfaffenhofen dataset， the model reached an OA of 98.65%， effectively mitigating the misclassification of suburban areas while maintaining clear edges for forests and open areas. Parameter analysis indicated that a 5% training sample ratio provides an optimal balance between classification accuracy and manual labeling costs. Furthermore， runtime analysis demonstrated that while the training time （505.30 s） is slightly higher due to the contrastive learning module， the inference speed （15.80 s） remains highly efficient for large-scale tasks.
Conclusion
2
This study presents a comprehensive end-to-end framework for refined PolSAR image classification by integrating region partitioning， heterogeneous feature learning， and differentiated contrastive enhancement. By utilizing the Hierarchical Semantic Model as a spatial guide， the network successfully overcomes the limitations of isotropic feature modeling. The dual-branch CNN-SGCN architecture ensures that fine-grained boundary details and global non-boundary structures are captured simultaneously. Moreover， the adaptive contrastive learning mechanism provides targeted supervision for difficult boundary pixels， significantly enhancing the discriminative power of the feature space. Experimental results across multiple bands （C and L） and varied scenes confirm that the proposed method achieves superior classification performance， boundary maintenance， and spatial consistency compared to existing state-of-the-art models. This research provides a robust and effective solution for the fine-grained analysis of complex heterogeneous PolSAR imagery.
An W ， Cui Y and Yang J . 2010 . Three-component model-based decomposition for polarimetric SAR data . IEEE Transactions on Geoscience and Remote Sensing ， 48 （ 6 ）： 2732 - 2739
Bi H X ， Kuang Z Z ， Li F ， Gao J H and Xu C . 2024 . Overview of deep learning algorithms for PolSAR image classification . Chinese Science Bulletin ， 69 （ 35 ）： 5107 - 5125
毕海霞 ， 况祖正 ， 李凡 ， 高静怀 ， 徐晨 . 2024 . 极化SAR图像分类深度学习算法综述 . 科学通报 ， 69 （ 35 ）： 5107 - 5125 ［ DOI： 10.1360/TB-2024-0087 http://dx.doi.org/10.1360/TB-2024-0087 ］
Cloude S R and Pottier E . 1997 . An entropy based classification scheme for land applications of polarimetric SAR . IEEE Transactions on Geoscience and Remote Sensing ， 35 （ 1 ）： 68 - 78
Cui Y H ， Liu F ， Jiao L C ， Guo Y W ， Liang X F ， Li L ， et al . 2021 . Polarimetric multipath convolutional neural network for PolSAR image classification . IEEE Transactions on Geoscience and Remote Sensing ， 60 ： 1 - 18 ［ DOI： 10.1109/TGRS.2021.3071559 http://dx.doi.org/10.1109/TGRS.2021.3071559 ］
Ding L . 2024 . Joint spatio-temporal modeling for semantic change detection in remote sensing images . IEEE Transactions on Geoscience and Remote Sensing ， 62 ： 1 - 14 ［ DOI： 10.1109/TGRS.2024.3362795 http://dx.doi.org/10.1109/TGRS.2024.3362795 ］
Freeman A and Durden S L . 1998 . A three-component scattering model for polarimetric SAR data . IEEE Transactions on Geoscience and Remote Sensing ， 36 （ 3 ）： 963 - 973
Gao H ， Wang C C ， Xiang D L ， Ye J W and Wang G Y . 2021 . TSPol-ASLIC： Adaptive superpixel generation with local iterative clustering for time-series quad- and dual-polarization SAR data . IEEE Transactions on Geoscience and Remote Sensing ， 60 ： 1 - 15 ［ DOI： 10.1109/TGRS.2021.3126669 http://dx.doi.org/10.1109/TGRS.2021.3126669 ］
Guan D ， Zhang L ， Wang Y and Liu J . 2024 . Multi-scale contrastive learning for few-shot PolSAR image classification . IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing ， 17 ： 1120 - 1135
He C ， Li Y ， Liao M and Yang J . 2020 . A novel PolSAR image classification method based on spatio-spectral feature fusion . IEEE Geoscience and Remote Sensing Letters ， 17 （ 1 ）： 92 - 96
He M Y ， Li Q ， Guo Y and Yan H M . 2025 . A survey and prospect of intelligent hyperspectral image classification . Journal of Image and Graphics ， 30 （ 6 ）： 2207 - 2238
何明宇 ， 李谦 ， 郭阳 ， 严鸿蒙 . 2025 . 智能高光谱图像分类研究进展与展望 . 中国图象图形学报 ， 30 （ 6 ）： 2207 - 2238 ［ DOI： 10.11834/jig.240532 http://dx.doi.org/10.11834/jig.240532 ］
Jiang N N ， Zhao W B ， Guo J ， Zhao Q and Zhu J . 2025 . Multi-scale feature extraction with 3D complex-valued network for PolSAR image classification . Remote Sensing ， 17 （ 15 ）： 2663 ［ DOI： 10.3390/rs17152663 http://dx.doi.org/10.3390/rs17152663 ］
Jiang X C and Zhang X W . 2025 . Multidimensional graph contrastive learning based on graph enhancement and multi-neural networks . Journal of Image and Graphics ， 30 （ 9 ）： 3097 - 3110
蒋晓灿 ， 张晓雯 . 2025 . 图增强与多神经网络结合的多维图对比学习 . 中国图象图形学报 ， 30 （ 9 ）： 3097 - 3110 ［ DOI： 10.11834/jig.240612 http://dx.doi.org/10.11834/jig.240612 ］
Kipf T N and Welling M . 2017 . Semi-supervised classification with graph convolutional networks // Proceedings of the 5th International Conference on Learning Representations . Toulon ： ICLR
Lee J S ， Grunes M R and Kwok R . 1999 . Classification of multi-look polarimetric SAR imagery based on complex Wishart distribution . International Journal of Remote Sensing ， 20 （ 8 ）： 1603 - 1618
Li M S ， Li W ， Liu Y K ， Huang Y W and Yang G P . 2023 . Adaptive mask sampling and manifold to Euclidean subspace learning with distance covariance representation for hyperspectral image classification . IEEE Transactions on Geoscience and Remote Sensing ， 61 ： 1 - 18 ［ DOI： 10.1109/TGRS.2023.3265388 http://dx.doi.org/10.1109/TGRS.2023.3265388 ］
Li Y ， Chen J and Liu L . 2023 . 3D-CNN with multi-scale feature fusion for polarimetric SAR image classification . Remote Sensing ， 15 （ 8 ）： 2145
Li Y ， Yang J F ， Zhang H S ， Li G and Chen J . 2024 . Review of polarimetric synthetic aperture radar remote sensing for land cover classification . National Remote Sensing Bulletin ， 28 （ 8 ）： 1835 - 1853
李煜 ， 杨静飞 ， 张鸿生 ， 李刚 ， 陈杰 . 2024 . 极化合成孔径雷达遥感地物分类研究进展 . 遥感学报 ， 28 （ 8 ）： 1835 - 1853 ［ DOI： 10.11834/jrs.20242346 http://dx.doi.org/10.11834/jrs.20242346 ］
Liu F ， Shi J F ， Jiao L C ， Liu H Y ， Yang S Y ， Wu J ， et al . 2016 . Hierarchical semantic model and scattering mechanism based PolSAR image classification . Pattern Recognition ， 59 ： 325 - 342 ［ DOI： 10.1016/j.patcog.2016.02.020 http://dx.doi.org/10.1016/j.patcog.2016.02.020 ］
Liu H Y ， Zhu T W ， Shang F H ， Liu Y Y ， Lv D R and Yang S Y . 2021 . Deep fuzzy graph convolutional networks for PolSAR imagery pixelwise classification . IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing ， 14 ： 504 - 514
Li W M ， Xia H ， Zhang J D ， Wang Y ， Jia Y and He Y H . 2024 . Complex-valued 2D-3D hybrid convolutional neural network with attention mechanism for PolSAR image classification . Remote Sensing ， 16 （ 16 ）： 2908 ［ DOI： 10.3390/rs16162908 http://dx.doi.org/10.3390/rs16162908 ］
Ren B ， Hou B ， Li M and Jiao L . 2022 . Contrastive learning-based polarimetric SAR image classification . IEEE Transactions on Geoscience and Remote Sensing ， 60 ： 1 - 14
Shi J F ， Ji S S ， Jin H Y ， Li J H ， Gong M G and Lin W S . 2025 . Content-adaptive multi-region deep network for polarimetric SAR image classification . IEEE Transactions on Circuits and Systems for Video Technology ， 35 （ 1 ）： 617 - 631 ［ DOI： 10.1109/TCSVT.2024.3456480 http://dx.doi.org/10.1109/TCSVT.2024.3456480 ］
Tian K ， Liu X C ， Tao D P and Ni J . 2025 . One unsupervised feature selection method for the classical linear classifier in land coverage classification with PolSAR imagery . Computational Intelligence ， 41 （ 1 ）： e 70025 ［ DOI： 10.1111/coin.70025 http://dx.doi.org/10.1111/coin.70025 ］
Wang S ， Ji K F and Dong K . 2023 . Polarimetric SAR land cover classification combining self-attention mechanism and graph convolutional networks . Journal of Image and Graphics ， 28 （ 4 ）： 1142 - 1155
王爽 ， 纪凯夫 ， 董凯 . 2023 . 自注意力机制与图卷积网络结合的极化SAR土地覆盖分类 . 中国图象图形学报 ， 28 （ 4 ）： 1142 - 1155 ［ DOI： 10.11834/jig.220256 http://dx.doi.org/10.11834/jig.220256 ］
Wang Z Q ， Li Y ， Zhang R ， Wang J B ， Li Y C and Chen Y . 2024 . A review of few-shot SAR image classification methods . Journal of Image and Graphics ， 29 （ 7 ）： 1902 - 1920
王智强 ， 李阳 ， 张睿 ， 王建波 ， 李宇超 ， 陈勇 . 2024 . 少样本SAR图像分类方法综述 . 中国图象图形学报 ， 29 （ 7 ）： 1902 - 1920 ［ DOI： 10.11834/jig.230563 http://dx.doi.org/10.11834/jig.230563 ］
Wu Y ， Ji K ， Yu W and Su Y . 2022 . Deep learning for polarimetric SAR image classification： a review of algorithms and targets . Remote Sensing ， 14 （ 14 ）： 3326
Wu Z Z ， Wan L ， Hong F H ， Tang Z D ， Sun F ， Zou L and Wang X F . 2025 . Graph convolutional network integrating skeleton large kernel operators and global context information . Journal of Image and Graphics ， 30 （ 11 ）： 3604 - 3616
吴志泽 ， 万龙 ， 洪芳华 ， 汤正道 ， 孙斐 ， 邹乐 ， 王晓峰 . 2025 . 融合骨架大核算子和全局上下文信息的图卷积网络 . 中国图象图形学报 ， 30 （ 11 ）： 3604 - 3616 ［ DOI： 10.11834/jig.240353 http://dx.doi.org/10.11834/jig.240353 ］
Zhang Q ， Xu F and Jin Y Q . 2024 . Superpixel-based random forest for high-resolution PolSAR image understanding . IEEE Transactions on Geoscience and Remote Sensing ， 62 ： 450 - 468
Zhang Z ， Guo H ， Yang J S ， Wang X G and Du Y . 2024 . Adversarial network with higher order potential conditional random field for PolSAR image classification . IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing ， 17 ： 1795 - 1808 ［ DOI： 10.1109/JSTARS.2023.3322344 http://dx.doi.org/10.1109/JSTARS.2023.3322344 ］
Zhu X X ， Tuia D ， Mou L C ， Xia G S ， Zhang L P ， Xu F ， et al . 2020 . Deep learning in remote sensing： a comprehensive review . IEEE Geoscience and Remote Sensing Magazine ， 8 （ 4 ）： 8 - 36
相关作者
相关机构
京公网安备11010802024621
