面向对象变化检测中多时相图像分割模式影响评价

Evaluating the effectiveness of multi-temporal image segmentation on object-based change detection

面向对象变化检测是高分辨率遥感图像分析技术中的研究热点,在国土资源监测、城市扩展、森林植被变化等方面具有广泛的应用前景.多时相图像分割是面向对象变化检测的关键步骤,主要包括三种模式:多时相组合分割、单时相分割与多时相分别分割.本文通过分析三种多时相图像分割模式下变化对象的差异,评价多时相图像分割策略对于面向对象变化检测结果的影响.结果表明,多时相分割模式对变化对象形状以及检测精度的影响均较大,三种模式中的多时相组合图像分割模式对本文研究区的变化检测精度最高.

Object-based image analysis（OBIA）has shown improved performances over the classical pixel-based methods,and object-based change detection（OBCD）is an important part of OBIA.From the perspective for object extraction of multi-temporal image data,the image segmentation can be categorized into three different models：two time data stacking as a whole for segmentation,which produce spatially corresponding objects;Extracting objects from one time data and assigning to the other time data without segmentation;segmenting independently for the two time data.The evaluation of three multi-temporal image segmentation models remains a critical significance because objects in different models are of various sizes and shapes.In this paper we use change vector analysis to divide objects into changed and unchanged objects,the changed objects acquired from the three image segmentation models are analyzed using qualitative and quantitative comparison,and the standard evaluation map is acquired by the artificial visual in-terpretation.From the comparison of three multi-temporal image segmentation models,it shows that the first image segmentation model has the highest overall accuracy and Kappa coefficient on both of the two study areas.In practical OBCD applications,we can choose the appropriate image segmentation models according to the status of study area and application purposes.