浏览全部资源
扫码关注微信
Published： 16 October 2023 ，
移动端阅览
张帆， 闫敏超， 倪军， 项德良. 2023. 高阶条件随机场引导的多分支极化SAR图像分类. 中国图象图形学报， 28(10):3267-3280
Zhang Fan， Yan Minchao， Ni Jun， Xiang Deliang. 2023. High-order conditional random fields-relevant multi-branch polarimetric SAR image classification. Journal of Image and Graphics， 28(10):3267-3280
目的
2
针对极化合成孔径雷达（polarimetric synthetic aperture radar，PolSAR）小样本分类问题，基于充分挖掘有限样本的极化、空间特征考虑，提出一种由高阶条件随机场（conditional random field，CRF）引导的多分支分类网络模型。
方法
2
利用Yamaguchi非相干目标分解方法，构建每个像素的极化特征向量。设计了由高阶CRF能量函数引导的多卷积分支特征提取网络，将像素点极化特征向量作为输入，分别提取像素点的像素特征、邻域特征和位置特征信息。将以上特征进行加和融合，并输入到 softmax 分类器中得到预分类结果。利用超像素方法对预分类结果图进行进一步修正和调优，平滑相邻像素之间的特异性和相似性。
结果
2
采用1%的采样率对两组真实的极化SAR数据进行测试。同时，为了更好地模拟实际应用中训练样本位置分布不均匀的情况，考虑了空间不相交采样方法作为对比实验。综合两种采样策略的实验结果表明，相较于只利用像素级特征或简单利用空间特征的方法，本文方法总分类精度平均提升7%～10%，不同地物类别的分类精准度均在90%以上，运行速度相比于支持向量机（support vector machine，SVM）提高了2.5倍以上。
结论
2
通过构建高阶CRF引导的卷积神经网络，将像素特征信息、同质区域特征和地理位置信息进行融合，有效建立了像素级和对象级数据之间的尺度关联，进一步扩充了像素点之间的空间依赖性，提取到了更强大更准确的表征特征，显著提高了标记样本数量较少情况下的卷积网络模型的分类性能，进一步保证了地物目标散射机制表征的全面性和可靠性。
Objective
2
Polarimetric synthetic aperture radar （PolSAR） is essential for high spatial resolution earth observation， and image classification can be as a key branch of PolSAR image interpretation. The emerging convolutional neural network （CNN） has its potentials in relevance to PolSAR image classification， but its accuracy and generalization ability is still challenged for its SAR labeling samples-derived constraints. We develop a multi-branch classification network model， which can integrate classification-refined location and semantic information of ground-based objects in polarimetric SAR images.
Method
2
First， polarimetric SAR data are analyzed and interpreted in terms of scattering model-based Yamaguchi four-component decomposition method. To extract spatial features at different levels further， channels-related concatenation are conducted through 1） the decomposed surface scattering， 2） double scattering， 3） volume scattering and 4） helix scattering. Second， to optimize pixels features of PolSAR images as classification objects， a high-order conditional random field （CRF） energy function-guided multi-branch CNN feature extraction model is designed to extract 1） pixel feature information， 2） azimuth correction neighborhood information， and 3）position coordinate information， which is used to describe the relationship between global spatial features and local features. We design the direction correction pixel block as well， which is different from the traditional two-dimensional matrix. To optimize the effect of different types of pixels on the center pixel and the classification further， error points in the neighborhood information can be modified， especially for edge pixels. Finally， superpixel constraint module-related adaptive polarization linear iterative clustering （Pol-ASLIC） method is used to generate a superpixel segmentation image. For each pixel in the small superpixel interval， the average probability of the pre-classification results is calculated， and the most probable class is assigned to each pixel as the total class. It can reduce classification-derived interference of speckle noise and smooth adjacent pixels-between heterogeneity and homogeneity， and such of classification results can be more compactable. In the experiment， the simulation of ground truth-related uneven spatial distribution of samples are carried out， and the spatial sampling-decoupled method is adopted as the comparative experiment of random sampling method. The spatial sampling-decoupled method can generate mutually independent training samples and minimize the interference of the sampling position.
Result
2
To alleviate the impact of network instability on the classification results， each experiment is replicated for 10 times， and the average value is taken as the final display result. Two groups of NASA/JPL AIRSAR system-acquired real polarimetric SAR images are tested with a sampling rate of 1%. Extensive qualitative and quantitative experimental results demonstrate that the method proposed can generate feasible analysis in terms of a small number of labeled samples. It can extract more comprehensive and effective features under different sampling strategies in comparison with the machine learning method and traditional convolutional classification model. Compared to methods using pixel-level features or spatial features only， overall classification accuracy is improved by an average of 7%～10%. The classification accuracy of each category in the random sampling method is reached above 90% on both datasets. Furthermore， running speed is 2.5 times faster than support vector machine（SVM）.
Conclusion
2
We develop a multi-branch convolutional neural network to extract more effective and accurate eigenvalues through pixel feature-fused information， homogeneous region features， and geographic location information. To strengthen scale correlations between pixel-level and object-level data effectively， pixels-between spatial feature correlation can be further extended， in which the classification performance of the CNN model can be significantly improved in terms of a small number of labeled samples. The potential accuracy of remote sensing data terrain classification can be predicted to preserve the comprehensiveness and reliability of the characterization of ground-based object scattering models.
SAR图像分类卷积神经网络（CNN）条件随机场（CRF）超像素分割采样策略Yamaguchi分解
SAR image classificationconvolutional neural network （CNN）conditional random field （CRF）superpixel segmentationsampling strategyYamaguchi decomposition
Chen S W and Tao C S. 2018. PolSAR image classification using polarimetric-feature-driven deep convolutional neural network. IEEE Geoscience and Remote Sensing Letters， 15（4）： 627-631 ［DOI： 10.1109/LGRS.2018.2799877http://dx.doi.org/10.1109/LGRS.2018.2799877］
Han W， Wang R S， Huang D Q and Xu C. 2020. Large-scale ALS data semantic classification integrating location-context-semantics cues by higher-order CRF. Sensors， 20（6）： #1700 ［DOI： 10.3390/s20061700http://dx.doi.org/10.3390/s20061700］
Hänsch R， Ley A and Hellwich O. 2017. Correct and still wrong： the relationship between sampling strategies and the estimation of the generalization error//2017 IEEE International Geoscience and Remote Sensing Symposium. Fort Worth， USA： IEEE： 3672-3675 ［DOI： 10.1109/IGARSS.2017.8127795http://dx.doi.org/10.1109/IGARSS.2017.8127795］
He K M and Sun J. 2015. Convolutional neural networks at constrained time cost//Proceedings of 2015 IEEE Conference on Computer Vision and Pattern Recognition. Boston， USA： IEEE： 5353-5360 ［DOI： 10.1109/CVPR.2015.7299173http://dx.doi.org/10.1109/CVPR.2015.7299173］
Hua W Q， Wang S， Guo Y H and Xie W. 2019. Semi-supervised PolSAR image classification based on the neighborhood minimum spanning tree. Journal of Radars， 8（4）： 458-470
滑文强， 王爽， 郭岩河， 谢雯. 2019. 基于邻域最小生成树的半监督极化SAR图像分类方法. 雷达学报， 8（4）： 458-470 ［DOI： 10.12000/JR18104http://dx.doi.org/10.12000/JR18104］
Lafferty J D， McCallum A and Pereira F C N. 2001. Conditional random fields： probabilistic models for segmenting and labeling sequence data//Proceedings of the 18th International Conference on Machine Learning. San Francisco， USA： Morgan Kaufmann Publishers： 282-289
Li Y F and Dong H B. 2018. Classification of remote-sensing image based on convolutional neural network. CAAI Transactions on Intelligent Systems， 13（4）： 550-556
李亚飞， 董红斌. 2018. 基于卷积神经网络的遥感图像分类研究. 智能系统学报， 13（4）： 550-556 ［DOI： 10.11992/tis.201706078http://dx.doi.org/10.11992/tis.201706078］
Liang J， Zhou J， Qian Y T， Wen L， Bai X and Gao Y S. 2017. On the sampling strategy for evaluation of spectral-spatial methods in hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing， 55（2）： 862-880 ［DOI： 10.1109/TGRS.2016.2616489http://dx.doi.org/10.1109/TGRS.2016.2616489］
Paoletti M E， Haut J M， Plaza J and Plaza A. 2019. Deep learning classifiers for hyperspectral imaging： a review. ISPRS Journal of Photogrammetry and Remote Sensing， 158： 279-317 ［DOI： 10.1016/j.isprsjprs.2019.09.006http://dx.doi.org/10.1016/j.isprsjprs.2019.09.006］
Tang L Y， Liu W， Yang D， Chen L， Su Y M and Xu X L. 2018. Flooding monitoring application based on the object-oriented method and Sentinel-1A SAR data. Journal of Geo-Information Science， 20（3）： 377-384
汤玲英， 刘雯， 杨东， 陈乐， 苏扬媚， 徐宪立. 2018. 基于面向对象方法的Sentinel-1A SAR在洪水监测中的应用. 地球信息科学学报， 20（3）： 377-384 ［DOI： 10.12082/dqxxkx.2018.170599http://dx.doi.org/10.12082/dqxxkx.2018.170599］
van der Maaten L and Hinton G. 2008. Visualizing data using t-SNE. Journal of Machine Learning Research， 9（86）： 2579-2605
Wang J， Zheng T， Lei P and Wei S M. 2018. Study on deep learning in radar. Journal of Radars， 7（4）： 395-411
王俊， 郑彤， 雷鹏， 魏少明. 2018. 深度学习在雷达中的研究综述. 雷达学报， 7（4）： 395-411 ［DOI： 10.12000/JR18040http://dx.doi.org/10.12000/JR18040］
Wang Y B， Mei J， Zhang L Q， Zhang B， Zhu P P， Li Y and Li X G. 2019. Self-supervised feature learning with CRF embedding for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing， 57（5）： 2628-2642 ［DOI： 10.1109/TGRS.2018.2875943http://dx.doi.org/10.1109/TGRS.2018.2875943］
Xiang D L， Ban Y F， Wang W and Su Y. 2017. Adaptive superpixel generation for polarimetric SAR images with local iterative clustering and SIRV model. IEEE Transactions on Geoscience and Remote Sensing， 55（6）： 3115-3131 ［DOI： 10.1109/TGRS.2017.2662010http://dx.doi.org/10.1109/TGRS.2017.2662010］
Yamaguchi Y， Moriyama T， Ishido M and Yamada H. 2005. Four-component scattering model for polarimetric SAR image decomposition. IEEE Transactions on Geoscience and Remote Sensing， 43（8）： 1699-1706 ［DOI： 10.1109/TGRS.2005.852084http://dx.doi.org/10.1109/TGRS.2005.852084］
Yang X F， Ye Y M， Li X T， Lau R Y K， Zhang X F and Huang X H. 2018. Hyperspectral image classification with deep learning models. IEEE Transactions on Geoscience and Remote Sensing， 56（9）： 5408-5423 ［DOI： 10.1109/TGRS.2018.2815613http://dx.doi.org/10.1109/TGRS.2018.2815613］
Zhang F， Ni J， Yin Q， Li W， Li Z， Liu Y F and Hong W. 2017. Nearest-regularized subspace classification for PolSAR imagery using polarimetric feature vector and spatial information. Remote Sensing， 9（11）： #1114 ［DOI： 10.3390/rs9111114http://dx.doi.org/10.3390/rs9111114］
Zhang F， Yan M C， Hu C， Ni J and Zhou Y S. 2021. Integrating coordinate features in CNN-based remote sensing imagery classification. IEEE Geoscience and Remote Sensing Letters， 19： #5502505 ［DOI： 10.1109/LGRS.2020.3045744http://dx.doi.org/10.1109/LGRS.2020.3045744］
Zhang H， Ye X， Wang C， Zhang B， Wu F and Tang Y X. 2014. The processing and applications of high resolution SAR images with object-based image analysis. Journal of Image and Graphics， 19（3）： 344-357
张红， 叶曦， 王超， 张波， 吴樊， 汤益先. 2014. 面向对象的高分辨率SAR图像处理及应用. 中国图象图形学报， 19（3）： 344-357 ［DOI： 10.11834/jig.20140302http://dx.doi.org/10.11834/jig.20140302］
Zhao F， Zhang W K， Yan Z Y， Yu H F and Diao W H. 2019. Multi-feature map pyramid fusion deep network for semantic segmentation on remote sensing data. Journal of Electronics and Information Technology， 41（10）： 2525-2531
赵斐， 张文凯， 闫志远， 于泓峰， 刁文辉. 2019. 基于多特征图金字塔融合深度网络的遥感图像语义分割. 电子与信息学报， 41（10）： 2525-2531 ［DOI： 10.11999/JEIT190047http://dx.doi.org/10.11999/JEIT190047］
Zhao Q H， Xie K L， Wang G H and Li Y. 2020. Land cover classification of polarimetric SAR with fully convolution network and conditional random field. Acta Geodaetica et Cartographica Sinica， 49（1）： 65-78
赵泉华， 谢凯浪， 王光辉， 李玉. 2020. 全卷积网络和条件随机场相结合的全极化SAR土地覆盖分类. 测绘学报， 49（1）： 65-78 ［DOI： 10.11947/j.AGCS.2020.20190038http://dx.doi.org/10.11947/j.AGCS.2020.20190038］
相关文章
相关作者
相关机构
京公网安备11010802024621