An approach to land use change detection by using high resolution remote sensing images is put forward in this paper. With the help of GIS land use map, image objects can be obtained by the matching of land use map and remote sensing images in the same region. Meanwhile pixel - based supervised classification is conducted for each image so that each pixel has its own class code. Then image subsegments can be obtained based on the image segment and the class code of each pixel within it. Image subsegments can be regarded as the basic units for feature extraction. Correlation coefficient is used for detecting changes between the images gotten from different time periods, and instead of the empirical selection, the change threshold is founded automatically by using ROC curve (receiver operating characteristic curve). Two multispectral Quickbird images obtained in 2002 and 2005 respectively and a 1 : 10 000 land use map of 2002 in the same region were used in the experiment. This study area is located in Wuhan City and the result shows that most land use changes can be detected, and hence this approach is effective.%提出一种利用高分辨率遥感影像进行土地利用变化检测的方法.以土地利用图为辅助数据,通过土地利用图和遥感影像的配准套合,获取影像像斑；同时,对遥感影像进行基于像素的监督分类,获取概略的类别图；再根据像斑内像素的类别编码完成子像斑的划分.以子像斑为影像分析的基本单位提取特征,以相关系数为相似性测度衡量不同时期子像斑的特征相似性,用ROC曲线(接受者操作特性曲线)代替经验选取的方法自动获取变化阈值,确定像斑是否发生变化.以武汉市区局部QuickBird 2002年和2005年多光谱影像、相同地区2002年1:10 000土地利用图为实验数据进行了算法的实验,结果显示绝大部分的变化区域都可以被提取出来,实验方法可行.
展开▼
