一种顾及背景信息的遥感影像自动化水体提取方法

An Automatic Water Extraction Method based on Remote Sensing Image with Background Information

水资源是地球生态系统的重要组成部分,准确有效的水体提取可为水文、气候建模、生态系统评估等不同学科研究提供基础性数据。针对遥感水体提取方法中存在的最优阈值难以确定、建筑阴影噪声难以抑制等问题,提出了一种基于遥感特征指数的水体提取方法,能够削弱指数法对最优阈值的依赖性并抑制建筑阴影产生的噪声。针对水体提取模型自动化程度不高、单一方法难以针对不同背景水体完成高质量提取等问题,发展了顾及背景信息的自动化水体提取方法,基于OTSU阈值法利用植被指数、建筑指数与坡度等信息对背景进行分类,然后采用不同的策略进行水体提取。实验证明,提出的方法自动化程度高,无需依赖先验知识,能在保证水体提取结果完整性的同时抑制由建筑与山体阴影产生的错误信息。

Water resources are an important part of the earth’s ecosystem.Accurate and effective water extraction can provide basic data for studies in different disciplines such as hydrology,climate modeling and ecosystem assessment.In view of the difficulty in determining the optimal threshold value and suppressing the noise of building shadow in the remote sensing water extraction method,a water extraction method based on remote sensing feature index is proposed,which can weaken the dependence of the index method on the optimal threshold value and suppress the noise generated by building shadow.For water body extraction model do not have a high level of automation,a single method is difficult to complete high quality water for different background extraction and other issues,the development of the background information of automatic water extraction method,based on OTSU threshold method using vegetation index,building index and slope information such as the classification to the background,and then adopt different strategies for water body extraction.Experiments show that this method is highly automated and does not need to rely on prior knowledge.It can ensure the integrity of water extraction results while suppressing the error information generated by the shadow of buildings and mountains.