Früher|Nächster

Multi-feature-based classification method using random forest and superpixels for polarimetric SAR images

  • role： First author第一作者
  • Affiliation：

    China Communications Construction Company Second Highway Consultants Co.Ltd, Wuhan 430056, China

  • Email：xu_qiao_cug@126.com
  • Introduction：徐乔，1991年生，男，工程师，研究方向为公路地质灾害遥感监测。E-mail：xu_qiao_cug@126.com
XU Qiao1,  
  • Affiliation：

    China Communications Construction Company Second Highway Consultants Co.Ltd, Wuhan 430056, China

ZHANG Xiao1,  
  • Affiliation：

    China Communications Construction Company Second Highway Consultants Co.Ltd, Wuhan 430056, China

YU Shaohuai1,  
  • Affiliation：

    Faculty of Information Engineering, China University of Geosciences(wuhan), Wuhan 430074, China

CHEN Qihao2,  
  • role： Corresponding author通信作者
  • Affiliation：

    Faculty of Information Engineering, China University of Geosciences(wuhan), Wuhan 430074, China

  • Email：liuxg318@163.com
  • Introduction：刘修国，1969年生，男，教授，研究方向为遥感影像信息提取与3S集成。E-mail：liuxg318@163.com
LIU Xiuguo2*

Resümee

The classification technique plays an important role in the analysis of polarimetric synthetic aperture radar (PolSAR) images. PolSAR image classification is widely used in information extraction and scene interpretation or is performed as a preprocessing step for further applications. However, speckle noise appears in PolSAR images because of the coherent interference of waves reflected from elementary scatters. Such inherent speckle noise degrades the classification performance and brings difficulty for PolSAR image classification. Therefore, a novel supervised multi-feature-based classification method was proposed in this study. This method combined polarimetric signature information and spatial context information based on the random forest model. First, a modified simple linear iterative clustering algorithm was utilized to generate superpixels as classification elements by using the Pauli RGB image, which helped reduce speckle noise interference. Second, a high-dimensional polarimetric SAR feature image was constructed by collecting various polarimetric signatures generated by polarimetric decomposition and algebra operations. Then, the random forest model was trained on the basis of the PolSAR feature image by using training samples, and the number of classification votes of each decision tree in the random forest for each pixel was counted to compute the class probabilities of the superpixels. Finally, a neighborhood function was defined to express the spatial relationship among adjacent superpixels quantitatively, and the class probabilities of the superpixels were recalculated by the predefined neighborhood function in a Probabilistic Label Relaxation (PLR) procedure to reduce the interference of speckle noise. The final classification result was obtained by the maximum a posteriori decision rule when the iteration of PLR was terminated. Comparative experiments using different RADARSAT-2 images were conducted to evaluate the validity and applicability of the proposed method. The proposed approach achieved the highest accuracy (94.39% on the Flevoland RADARSAT-2 image and 85.09% on the Wuhan RADARSAT-2 image) and generated accurate and consistent classification results for the experimental images, which was considerably improved compared with those of other methods. Therefore, the proposed method can effectively suppress the interference of speckle noise by using superpixels and spatial context information and obtain accurate and consistent classification results for PolSAR images.

Schlüsselwort

remote sensing;polarimetric SAR;classification;multifeature;Random Forest;superpixels;probabilistic label relaxation

References

  1. 1.
    Achanta R, Shaji A, Smith K, Lucchi A, Fua P and Süsstrunk S. 2012. SLIC superpixels compared to state-of-the-art superpixel methods. IEEE Transactions on Pattern Analysis and Machine Intelligence, 34(11): 2274–2282
  2. 2.
    Akbari V, Doulgeris A P, Moser G, Eltoft T, Anfinsen S N and Serpico S B. 2013. A textural–contextual model for unsupervised segmentation of multipolarization synthetic aperture radar images. IEEE Transactions on Geoscience and Remote Sensing, 51(4): 2442–2453
  3. 3.
    Benz U and Pottier E. 2001. Object based analysis of polarimetric sar data in alpha-entropy-anisotropy decomposition using fuzzy classification by ecognition//Proceedings of International Geoscience and Remote Sensing Symposium. Sydney, NSW, Australia: IEEE: 1427-1429 [DOI: 10.1109/IGARSS.2001.976867]
  4. 4.
    Blaschke T. 2010. Object based image analysis for remote sensing. ISPRS Journal of Photogrammetry and Remote Sensing, 65(1): 2–16
  5. 5.
    Breiman L. 2001. Random forests. Machine Learning, 45(1): 5–32
  6. 6.
    Chen L J, Yang W, Liu Y and Sun H. 2010. Feature evaluation and selection for polarimetric SAR image classification//Proceedings of the IEEE 10th International Conference on Signal. Beijing, China: IEEE: 2202-2205 [DOI: 10.1109/ICOSP.2010.5656765]
  7. 7.
    Feng J L, Cao Z J and Pi Y M. 2014. Polarimetric contextual classification of PolSAR images using sparse representation and superpixels. Remote Sensing, 6(8): 7158–7181
  8. 8.
    Haddadi G A, Sahebi M R and Mansourian A. 2011. Polarimetric SAR feature selection using a genetic algorithm. Canadian Journal of Remote Sensing, 37(1): 27–36
  9. 9.
    Lee J S, Grunes M R and Pottier E. 2001. Quantitative comparison of classification capability: fully polarimetric versus dual and single-polarization SAR. IEEE Transactions on Geoscience and Remote Sensing, 39(11): 2343–2351
  10. 10.
    Liu B, Hu H, Wang H Y, Wang K Z, Liu X Z and Yu W X. 2013. Superpixel-based classification with an adaptive number of classes for polarimetric SAR images. IEEE Transactions on Geoscience and Remote Sensing, 51(2): 907–924
  11. 11.
    Liu H Y, Zhu D X, Yang S Y, Hou B, Gou S P, Xiong T and Jiao L C. 2016. Semisupervised feature extraction with neighborhood constraints for polarimetric SAR classification. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 9(7): 3001–3015
  12. 12.
    Maghsoudi Y, Collins M J and Leckie D G. 2013. RADARSAT-2 polarimetric SAR data for boreal forest classification using SVM and a wrapper feature selector. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 6(3): 1531–1538
  13. 13.
    Niu X and Ban Y F. 2012. An adaptive contextual sem algorithm for urban land cover mapping using multitemporal high-resolution polarimetric SAR data. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 5(4): 1129–1139
  14. 14.
    Qin F C, Guo J M and Lang F K. 2015. Superpixel segmentation for polarimetric SAR imagery using local iterative clustering. IEEE Geoscience and Remote Sensing Letters, 12(1): 13–17
  15. 15.
    Reigber A, Jäger M, Neumann M and Ferro-Famil L. 2010. Classifying polarimetric SAR data by combining expectation methods with spatial context. International Journal of Remote Sensing, 31(3): 727–744
  16. 16.
    Richards J A and Jia X P. 2007. A dempster-shafer relaxation approach to context classification. IEEE Transactions on Geoscience and Remote Sensing, 45(5): 1422–1431
  17. 17.
    Song H, Yang W, Bai Y and Xu X. 2015. Unsupervised classification of polarimetric SAR imagery using large-scale spectral clustering with spatial constraints. International Journal of Remote Sensing, 36(11): 2816–2830
  18. 18.
    Sun X, Huang P P, Tu S T and Yang X L. 2016. Polarimetric SAR image classification using multiple-feature fusion and ensemble learning. Journal of Radars, 5(6): 692–700
  19. 19.
    Wang W S, Yang X F, Li X F, Chen K S, Liu G H, Li Z W and Gade M. 2017. A fully polarimetric sar imagery classification scheme for mud and sand flats in intertidal zones. IEEE Transactions on Geoscience and Remote Sensing, 55(3): 1734–1742
  20. 20.
    Wu Y H, Ji K F, Yu W X and Su Y. 2008. Region-based classification of polarimetric SAR images using Wishart MRF. IEEE Geoscience and Remote Sensing Letters, 5(4): 668–672

Lesen Sie die ganze Passage

The above content is generated by Large Model Translation. The translated content is for reference only. We do not assume any commercial or legal responsibilty for any consequences arising from the use of our website