基于上下文相关和模糊聚类的无监督图像变化检测
首发时间:2017-04-26
摘要:提出了一种基于模糊聚类的无监督二时刻变化检测方法,适合于高分辨率二时刻图像和多光谱遥感图像的变化检测问题。不同于经典无监督方法中忽略像素相邻信息并假设类间边界清晰的前提,该方法不仅考虑由于类的重叠而导致的不清晰,并定义了软边界,还利用了局部的时间和空间的关联信息。实验结果表明,该方法达到了很高的变化检测精确度。
For information in English, please click here
Context-Sensitive Correlation based and Fuzzy Collaborative Clustering based Unsupervised Change Detection
Abstract:In this paper, a framework for unsupervised detection of binary change using fuzzy clustering is proposed for the high resolution bi-temporal and multispectral remote sensing images. Instead of neglecting the local neightborhood information and assuming the crisp boundaries between chaged and unchaged class in standard unsupervised methods, the proposed method not only takes care of ambiguity arises from overlapping changed and unchanged class by defining soft boundaries between them, but also incorporates local temporal and spatial associations. Experimental results show that proposed method achieve high change detection accuracy.
Keywords: context fuzzy clustering unsupervised change detection
论文图表:
引用
No.4725055119291014****
同行评议
共计0人参与
勘误表
基于上下文相关和模糊聚类的无监督图像变化检测
评论
全部评论0/1000