【摘要】 水体作为地表覆盖的重要组成之一,水体信息提取通常也是电子地图更新的首要环节。本文基于北京二号高分辨率遥感数据,采用面向对象的技术流程实现了对影像的多尺度分割。同时对全局最优阈值算法进行改进,对水体指数、植被指数等关键特征的取值进行自动计算,减少人工参与。最终提取的水体数量较为准确,边界规整,总体精度可以达到96%,Kappa系数为0.73,有效提升了信息提取的工作效率与质量。更多还原

【Abstract】 As one of the most significant steps, the extraction of water body information is the first link in the procedure for updating electronic maps. In this paper, with the pretreatment of TripleSat constellation remote sensing data and object-oriented technique process, the extraction rule set of water body was carried out based on the multiscale segmentation. At the same time, this paper improves the global optimal threshold algorithm and automatically calculates the values of key features such as water body index and vegetation index to reduce artificial participation.The final extraction result of water body was proved to be accurate with a regular shape, reaching an overall accuracy of 96% and Kappa coefficient of 0.73. The research in this paper reduces the human involvement greatly, contributing to improve the working efficiency and accuracy of water body extraction.更多还原