基于全极化SAR影像的双台河口湿地分类及其变化分析

为了研究辽东湾双台河口湿地变化原因，对该地区全极化SAR影像进行湿地分类.引入Freeman分解建模全极化SAR影像，得到二面角散射、体散射、表面散射的散射功率，利用SVM分类法对散射机制假彩色影像进行分类，进而确定该地区的湿地分布情况.通过对比分析2007年与2016年的湿地分布情况，同时考虑区域特性和数据的可获取性，选取盘锦市年降水量、年均气温、年径流量、海平面高度、城市建成面积、原油年产量、水产品年产量和逐年GDP等8个驱动因子，研究双台河口湿地变化的驱动机制.结果表明:①全极化SAR影像分类结果总精度达到80.25%，较光学影像的分类精度提升7.43%，并且可将芦苇湿地进一步细分为芦苇池塘和芦苇草甸；②2016年双台河口各湿地面积占比依次为草本沼泽（28.54%）>浅海水域（22.44%）>灌丛沼泽（16.88%）>水产养殖厂（9.54%）>河流（8.91%）>稻田（7.58%）>淤泥质沙滩（6.11%）；③不同湿地的变化原因有所差异，驱动因子也不尽相同，如水产品年产量增加是水产养殖厂转化为淤泥质沙滩的直接驱动因子，而气温上升等自然因素和城市建成区面积增长等社会因素是草本沼泽退化和变化的主要驱动因子.研究显示，全极化SAR影像较光学影像更适合湿地分类，而双台河口湿地自然湿地减少和人工湿地增加是自然和社会因素共同作用的结果. 

Full-PolSAR Image based on Wetland Classification and Change Analysis of the Shuangtai Estuary Wetland

To study changes of the Shuangtai Estuary Wetland, full-PolSAR image of the region was classified. Scattering powers of three scattering mechanisms (double-bounce scatter, volume scatter, surface scatter) were obtained by the Freeman decomposition and visualized as false color images. The false color images were classified by SVM classifier to determine the distribution of wetlands in the Shuangtai estuary area. Considering the relevance to wetland change and the data availability, we selected 8 driving factors (annual precipitation, average annual temperature, annual runoff, sea level, urban built-up area, annual crude oil production, annual output aquatic products and GDP of Panjin) to study the mechanisms of Shuangtai estuary wetland changes. As a result: (1) The total classification accuracy is 80.25% which is higher than the Landsat-8 image 7.43%. Meanwhile, full-PolSAR image can be used to obtain better classification results based on the differences of the scatter mechanisms among all the objects derived from the Freeman decomposition and the proposed method can not only classify wetland from full-PolSAR image but also subtly sub-classify the reed wetland into meadow reed and reed pond to comprehensively reflect the spatial distribution of various wetlands. (2) The area ratios for marsh, shallow sea waters, shrub swamp, reservoir pond, rivers, rice and silt sandy beaches paddies were 28.54%, 22.44%, 16.88%, 9.54%, 8.91%, 7.58%, 6.11%. (3) The results demonstrated that the main driving factors of the different types of wetland varied. For example, the increase in the annual output of aquatic products was a direct driving factor for the conversion of ponds to silt sandy beaches; natural factors such as rising temperatures and social factors such as increase in the urban development were the main driving factors for the degradation and changes of marshes. The study shows that the full-PolSAR image is more suitable for wetland classification than optical image. The natural and social factors are the main reasons of the Shuangtai Estuary wetland changes.