中文摘要：

该文旨在探讨中分辨率遥感影像的土地覆盖变化对象识别方法.首先基于植被-不透水层-土壤（VIS）模型,参考OIF指数的信息量进行对象特征指标选取,基于过分割和欠分割指数确定最优分割尺度,对两期SPOT影像进行多时相分割,然后利用卡方变换方法自动选择阈值,进行植被、不透水层、土壤和水体4种土地覆盖类型之间的变化对象识别.精度评价表明,随着对象特征指标包含信息量的不断增加,检测结果的总精度不断提高,其中对影像所有特征指标进行主成分分析并选择前3个主成分作为特征指标组合对土地覆盖变化对象进行识别的总精度最高,为93.9％,Kappa系数为0.824,证明了该方法的有效性.

英文摘要：

The aim of this study is to propose a methodology to detect change objects in remote sensing images with medium spatial resolution. The optimal combination of change feature indicators was chosen by Optimal Indicator Factor （OIF） based on the Vegetation-Impervious-Soil model at first. Then the multi-temporal segmentation is applied based on the selected segmentation scales chosen by the over- and undersegmentation index. Finally, the change object detection between vegetation, impervious, soil and water is proposed based on Chi-square transformation, an automated thresholding method. According to the accuracy assessment,with the increasing of information in different change indicators, the overall accuracy is also increasing, and the result based on the principal component analysis has the highest overall accuracy of 93.9% with the Kappa of 0. 824. The result indicates that the Chi-square transformation method is very effective with medium spatial resolution images.