基于Landsat8 OLI影像的哈尔滨市辖区水体信息提取方法对比分析

Comparative Analysis of Water Information Extraction Methods in Harbin Area Based on Landsat8 OLI Images

以Landsat8 OLI影像为数据源,以哈尔滨市辖区为研究范围,采用阈值法、NDWI指数法、MNDWI指数法以及EWI指数法,对研究区内水体信息进行提取.以人机交互结果作为研究区内水体信息真实值,通过查全率、查准率以及虚警率三个指标对提取结果进行精度评价.结果表明,四种提取方法都能够将大面积水体信息完整提取出来,但是在虚警率以及信息的细节处理上,EWI指数法以及MNDWI指数法在查全率达到0.9以上的同时,能够将虚警率保持在0.1以下.EWI指数法,在对细小河流的提取方面,以及降低居民地等地物的噪声影响方面在几种算法中表现最为优异,是最适合该研究内的水体提取算法.

Taking Landsat 8 OLI image as data source and Harbin Municipality as research area, using threshold method, NDWI index method, MNDWI index method and EWI index method, water information in the study area is extracted. The human-computer interaction results are used as the real value of water information in the study area. The accuracy of the extracted results is evaluated by recall rate, precision rate and false alarm rate. The results show that all four methods can extract large area water information completely, but in terms of false alarm rate and details of information processing, EWI index method and MNDWI index method can keep the false alarm rate below 0.1 while the recall rate is above 0.9. EWI index method is the most suitable water extraction algorithm in this study. It has the best performance in extracting small rivers and reducing the noise impact of residential areas.