The Influence of Polarimetric Parameters and an Object-Based Approach on Land Cover Classification in Coastal Wetlands

Remote Sensing

Yuanyuan Chen²  Xiufeng He²  Jing Wang¹  Ruya Xiao²  Alisa L. Gallant³  Nicolas Baghdadi³ 

[1] China Land Surveying and Planning Institute, Beijing 100035, China; E-Mail:;Institute of Satellite Navigation and Spatial Information System, Hohai University, Nanjing 210098, China; E-Mails:Institute of Satellite Navigation and Spatial Information System, Hohai University, Nanjing 210098, China;

关键词: wetlands;    ALOS PALSAR;    polarimetric decomposition;    object-based approach;    decision tree;   

DOI  :  10.3390/rs61212575

来源: mdpi

【 摘 要 】

The purpose of this study was to examine how different polarimetric parameters and an object-based approach influence the classification results of various land use/land cover types using fully polarimetric ALOS PALSAR data over coastal wetlands in Yancheng, China. To verify the efficiency of the proposed method, five other classifications (the Wishart supervised classification, the proposed method without polarimetric parameters, the proposed method without an object-based analysis, the proposed method without textural and geometric information and the proposed method using the nearest-neighbor classifier) were applied for comparison. The results indicated that some polarimetric parameters, such as Shannon entropy, Krogager_Kd, Alpha, HAAlpha_T11, VanZyl3_Vol, Derd, Barnes2_T33, polarization fraction, Barnes1_T33, Neuman_delta_mod and entropy, greatly improved the classification results. The shape index was a useful feature in distinguishing fish ponds and rivers. The distance to the sea can be regarded as an important factor in reducing the confusion between herbaceous wetland vegetation and grasslands. Furthermore, the decision tree algorithm increased the overall accuracy by 6.8% compared with the nearest neighbor classifier. This research demonstrated that different polarimetric parameters and the object-based approach significantly improved the performance of land cover classification in coastal wetlands using ALOS PALSAR data.

【 授权许可】

CC BY   
© 2014 by the authors; licensee MDPI, Basel, Switzerland.