中文摘要：

在光谱直接比较的变化向量分析基础上,提出一种利用景观类型的尺度响应特征进行城市扩张的变化检测方法,该方法充分挖掘了对象的几何特征与纹理特征。首先,利用基准期与检测期影像中不同景观对象的各自拟合特征对数据进行分割,获取子对象,根据每个对象内象元光谱具有同质性的原则,对景观类型发生变化的对象通过规则进行识别,以初步提取变化范围。进而,利用ROC曲线确定最佳分界点与可疑区间范围,以对象为基本单元采用模糊函数的思想对变化区域进行精确提取。最后,以山地城市——贵阳市主城区Landsat-5 2001年和2008年的多光谱影像为数据源,对方法进行了检验。结果表明,该方法可以准确、迅速的提取城市变化范围。

英文摘要：

A use of the scale of the landscape type response characteristics of urban expansion change detection method is proposed based on direct comparison of the changes in the spectral vector analysis. This method had full used geometry and texture features. First,the use of different landscape objects contemplated and features in the base period and the detection of image segmentation,cross-nested and presplits get the child object,and then according to the homogeneity of the spectral clustering standard initial extraction range. Finally,use the ROC curve to determine the best cut-off point and suspicious interval range,base on member ship and the basic unit of change vector analysis method for accurate extraction of the change detection area. Use the Landsat- 5 2001 and 2008multi-spectral image of the main city in Guiyang City,Guizhou Province of China as data source,to do experiments to prove the feasibility of the method. The result show this method can exactly and rapidly find urban of change.