基于改进标记分水岭分割的城市水体提取研究  

Study on urban water information extraction based on improved marked-based watershed image segmentation

作　　者：叶刚[1] YE Gang(Zhengzhou Urban Planning Design&Survey Research Institute,Zhengzhou 450000,China)

机构地区：[1]郑州市规划勘测设计研究院,河南郑州450000

出　　处：《人民长江》2020年第7期231-235,共5页Yangtze River

基　　金：国家自然科学基金项目(41901285)。

摘　　要：针对高分辨率遥感影像水体提取中光谱混淆、噪声干扰等问题,提出了一种改进标记分水岭的影像分割方法。该方法结合水体的光谱与几何特征建立模型提取初始水体信息,并利用标记分水岭算法实现水体区域的初始分割;然后计算各标记区域内的灰度均值,选取种子像斑样本,并以像斑为单元在梯度影像中进行区域生长;最后利用水体指数模型提取水体信息。以武汉市局部卫星影像为对象,提取水体信息。研究表明:改进的基于像斑区域生长的算法能够兼顾邻近区域内像素的几何结构信息,显著提高算法精度,该方法有着较高的处理效率和准确性,具有应用价值。To solve the problems of spectral confusion and noise interference in water information extraction from high-resolution remote sensing images,this paper proposed an improved marked-based watershed image segmentation method.The basic steps of the method included:(1)The initial segmentation results of the water region could be realized by marked-based watershed image segmentation,combined with the initial water-information established by the model of the spectral and geometric characteristics of the water.(2)The average gray in the marked region was calculated,and region growing was conducted in the gradient image through selecting seed image spot.(3)The final water extraction could be carried out by the water index model(normalized difference water index,NDWI).Taking the local satellite image of Wuhan as an object,the improved spot-based region growing algorithm in this paper takes account of the geometric structure information of pixels in the neighboring area,and significantly improves the accuracy of the algorithm.Experimental results demonstrate that this method has a high precision,rational robustness and practical value.

关 键 词：水体信息提取 标记分水岭分割 影像分割 像斑 

分 类 号：TP751.2[自动化与计算机技术—检测技术与自动化装置]