面向对象的土地利用变化检测方法研究

The Study on Land Use Change Detection Based on Object-oriented Analysis

利用分类后比较法进行土地利用变化检测时,常用的自动分类方法只能利用遥感数据的光谱信息,分类精度较低,本文在此基础上将面向对象的分类方法引入到变化检测中。该方法综合利用遥感数据光谱信息、纹理特征、拓扑关系和专题信息,在进行多尺度分割获取对象后的基础上,通过对对象的目视识别选择样本来进行分类。利用面向对象的方法成功检测出了所选取的试验区十年间的土地利用变化信息,得到了较为满意的结果,为土地资源可持续利用提供了依据。

Conventional map-to-map comparison method is used frequently to land use change detection from remote sensing data. Since only the spectral information is used in the process of classification, so its accuracy is low. This paper applies object- oriented image analysis to the change detection and can overcome the limitation stated above. Synthesizing spectral information, texture information, topological relation and thematic information, we get the classification object through multi-scale image segmentation techniques and then classify the image by selecting samples visually. In the study, it detects the land use change information of the test area during ten years through the method based on object-oriented analysis. The result is satisfying and can provide the foundation for sustainable development of land resource.