基于多层马尔科夫随机场模型的变化检测

Change Detection Based on Multilayer Markov Random Field Model

提出了一种基于多层马尔科夫随机场模型的变化检测方法.首先输入的多个特征通过模拟退火方法得到对应每个特征的初步检测结果,然后利用多层马尔科夫随机场将初步检测结果进行融合.最后对图像进行目标级分割并与像素级变化检测结果相结合,根据各个目标的变化率给出了最终的变化检测.文章提出的方法运用多层马尔科夫随机场融合方法充分考虑了相邻像素间的相关性和各层之间的联系,并从目标角度进行了变化率的分析,使得融合结果更细致和精确.

This paper proposes a change detection method based on Multilayer Markov random field. First of all,we need get some detection result using features selected. Next,Multilayer Markov random field is applied to fuse detection result from previous steps. Since Markov random field fusion method takes full account of the correlation between the adjacent pixels and the links of different layer,the fusion results are more accurate and practical.