基于GF-1影像的山区河道丰、枯水期水体提取方法改进  

Improving Water Body Extraction Method in a Mountainous River in Wet and Dry Seasons Based on GF-1 Images

作　　者：赵程铭 董晓华[1,2,3] 薄会娟 章程焱[1,2,3] 张庆玉 ZHAO Cheng-ming;DONG Xiao-hua;BO Hui-juan;ZHANG Cheng-yan;ZHANG Qing-yu(School of Water Conservancy and Environment,China Three Gorges University,Yichang 443002,Hubei,China;Engineering Research Center of the Ministry of Education for the Ecological Environment of the Three Gorges Reservoir Area,Yichang 443002,Hubei,China;Hubei Collaborative Innovation Center for Water Resources Security,Wuhan,Hubei 430072,China)

机构地区：[1]三峡大学水利与环境学院,湖北宜昌443002 [2]三峡库区生态环境教育部工程研究中心,湖北宜昌443002 [3]水资源安全保障湖北省协同创新中心,武汉430072

出　　处：《中国农村水利水电》2022年第9期155-161,共7页China Rural Water and Hydropower

基　　金：欧洲空间局、中国国家遥感中心“龙计划”5项目(58516)。

摘　　要：丰、枯水期山区下垫面地物特征如植被覆盖度呈显著季节性变化,因此同种水体提取方法在应用于不同水期会获得不同的提取效果;同时,山区河道水体位于峡谷内,易受山体阴影的影响,而且河道狭长细小,水体提取结果容易出现断线现象。为解决上述问题,本文使用国产GF-1遥感影像,首先对比单波段阈值法、NDWI水体指数和SWI阴影水体指数在丰、枯水期典型月对山区河道水体提取的应用效果,分别确定在丰、枯水期最适用的提取方法。然后以DEM数据为基础,考虑遥感成像时太阳高度与辐射角度,生成山体阴影图层,与前述方法获得的水体叠加,消除山体阴影的影响;然后再次以DEM数据为基础,提取河网水系,与遥感提取的水体叠加,以消除零散水体小斑块(如堰塘等)的影响。最后应用膨胀滤波和Pavlidis异步细化算法填充水体断线处。结果表明,丰、枯水期典型月水体提取的最适方法分别为单波段阈值法、SWI阴影水体指数法。通过山体阴影消除、零散水体消除、应用膨胀滤波和Pavlidis异步细化算法进行水体断线填充等综合改进处理后,丰、枯水期水体提取结果总体精度分别达到99.52%、99.27%,Kappa系数分别为0.98、0.97,精度达到“极好”的最优等级标准。本文的方法能有效地解决丰、枯水期在不同地物特征下山区河道水体提取存在的瓶颈问题,为后续应用遥感技术获取水体其他信息(如污染物浓度)奠定了良好基础。The characteristics of the underlying surface such as vegetation coverage showed significant seasonal changes during the wet and dry seasons,The same water extraction method can obtain different extraction results in different seasons;At the same time,the mountain water body is located in the valley,easily affected by the mountain shadow,and the river water body is long and narrow,easy to appear the water body line breaking phenomenon.In order to solve the above problems,China madeGF-1 remote sensing images were used in this paper.Firstly,the application effects of single-band threshold method,Normalized Difference Water Index and Shadow Water Index on mountain channel water extraction in typical months of wet and dry seasons were compared,and the most applicable extraction methods were determined respectively in wet and dry seasons.Then,based on DEM data,considering the sun height and radiation angle in remote sensing image,the mountain shadow layer is generated,which is superimposed with the water body obtained by the above method to eliminate the influence of mountain shadow;Then,based on the DEM data,the river network is extracted and superimposed with the water extracted by remote sensing to eliminate the influence of small patches of scattered water(such as ponds).Finally,the dilation filtering and image thinning are applied to fill the broken line.The results show that the single band threshold method and Shadow Water Index method are suitable in wet and dry seasons.Through the removal of mountain shadow,the elimination of scattered water bodies,the application of dilation filtering and image thinning to fill the broken lines of water bodies,the overall accuracy of water extraction in the wet and dry seasons respectively reached 99.52% and 99.27%,the Kappa coefficients were 0.98 and 0.97 respectively,Accuracy up to the optimal grade standard of "excellent".This method can effectively solve the bottleneck problem of river water extraction in mountainous area under different extraction background in wet and

关 键 词：高分一号 水体提取 山体阴影 膨胀滤波 图像细化 

分 类 号：P332[天文地球—水文科学] P237[水利工程—水文学及水资源]