基于Landsat 8 OLI的合肥市水域面积提取方法有效性研究

Study on the Validity of Water Area Extraction Method in Hefei Based on Landsat 8 OLI

目的:从亚像元、像元和对象的角度对不同类型水体信息进行识别提取,探讨Landsat 8 OLI数据对水体面积提取方法的有效性。方法:采用监督分类-神经网络法、面向对象法、MNDWIOSTU阈值法和线性混合像元分解法,分别对合肥市水体信息进行提取分析。结果:采用的4种方法均适用于合肥境内的水体研究,其中线性混合像元分解法提取精度最高且具有较强的鲁棒性,总体精度和Kappa系数的平均值高达97.05%和0.97。结论:针对中等分辨率的Landsat 8 OLI数据,基于亚像元的线性混合像元分解方法对水体识别具有较好的精度,可为更精细或更大尺度的水体信息提取提供技术参考。

Objective:Recognition and extraction of water information from the perspective of sub-pixel,pixel and object was used to explore the validity of water area extraction method based on optical remote sensing technology.Methods:The supervised classification-neural network method,object-oriented method,MNDWI threshold method,binary method and linear mixed pixel decomposition method were used to deal with the research areas in Hefei.Results:The four methods were suitable for the research area of water bodies containing different ground objects in Hefei.The linear mixed pixel decomposition method had high and stable extraction accuracy among the four classification methods,and the lowest mapping accuracy and overall accuracy were 92.86%and 96.22%.Conclusion:The linear mixed pixel decomposition method had good accuracy in extracting water body area,and this method can provide technical reference for water body information extraction.