基于高分二号影像的高潜水位矿区水体提取方法

Extraction method of high dive site mining area Water-body based on GF-2 data

淮南矿区位处高潜水位平原区,区域内水系发达,且长期地下煤炭开采使地表形成大面积水区。文中基于国产高分二号(GF-2)影像数据,以淮南潘谢矿区三座矿山为例,利用2016年GF-2多光谱与全色融合数据,选择单波段阈值法、多波段谱间关系法、归一化水体指数法(NDWI)三种水体信息自动提取方法和监督分类法,对研究区内水体进行提取。结合实地采样得到的混淆矩阵对水体区域的整体提取结果进行精度验证对比分析。实验结果表明:利用监督分类法提取水体信息效果最好,总体精度达到98.18%,Kappa系数为0.94;自动提取方法中单波段阈值法效果最好,总体精为93.26%,Kappa系数为0.81。本研究对比分析得基于GF-2遥感数据的最优矿区水体信息提取方法为监督分类法,为后期矿区水体的动态变化预测和治理提供了科学依据。

Huainan mining area is located in the high water level plain area,the area of water system developed,and a large area of surface water area was formed because of the long-term underground coal mining. Water body information was extracted based on GF-2 data. Taking the GF-2 image data of 2016 selecting three kinds of water body information automatic extraction method including the single-band threshold method,the band relation method and the normalized water index method（ NDWI） and supervised classification taking the three mines in Panzhi Mining Area in Huainan as an example in the study area. Combing with the confusion matrix obtained from field sampling,the overall results of water body extraction were compared and verified with accuracy. Experimental results showed that the effect of using supervised classification was the best. The overall accuracy was 98. 18% and the Kappa coefficient was 0. 94. And the effect of single-band threshold method was best in three automatic extraction methods. The total precision was 93. 26% and the Kappa coefficient was 0. 81. In this paper,the supervised classification method was the best extraction method of water body information in mining area,which provided a scientific basis for the prediction and control of the dynamic change of the water body in the later mining area.