上海交通大学学报(自然版) ›› 2012, Vol. 46 ›› Issue (02): 245-249.

• 自动化技术、计算机技术 • 上一篇    下一篇

基于三维直方图聚类特性的非监督图像变化检测方法  

孙文邦1,2,唐海燕1,陈贺新2,吴迪1   

  1. (1空军航空大学 特种专业系, 长春 130022; 2吉林大学 通信工程学院, 长春 130025)   Key words: change detection; 3D histogram; clustering characteristic
  • 收稿日期:2010-12-08 出版日期:2012-02-28 发布日期:2012-02-28
  • 基金资助:

    国家自然科学基金国际合作项目(609111301281),国家自然科学基金重点项目(60832002)

Research of Unsupervised Change Detection Method Based on the Clustering Characteristic of 3D Histogram

 SUN  Wen-Bang-1, 2 , TANG  Hai-Yan-1, CHEN  He-Xin-2, WU  Di-1   

  1. (1. Aviation Information Department, Aviation University of Air Force, Changchun 130022, China;2. School of Communication Engineering, Jilin University, Changchun 130025, China)
  • Received:2010-12-08 Online:2012-02-28 Published:2012-02-28

摘要: 针对图像非监督变化检测算法进行了研究,提出了一种新的检测方法. 基于差异图像的三维直方图聚类特性,利用改进的一维最大熵寻找三维直方图最佳的平面阈值和平面方向;由其共同确定三维直方图的最佳划分平面,将三维直方图划分成变化聚类和非变化聚类;再根据三维直方图聚类划分结果完成差异图像的变化检测. 理论分析和实验结果表明,该方法相对传统检测方法具有更好的检测性能.    

关键词: 变化检测, 三维直方图, 聚类特性

Abstract: The algorithm of image change detection was researched, and a novel unsupervised change detection technique was proposed. First, the optimal plane threshold and the optimal plane direction are searched by using improved 1D maximal entropy principle based on the 3D histogram clustering characteristic of different image. Then the segmentation plane is established to segment the 3D histogram into changed cluster and unchanged cluster. Finally, the changed pixels and unchanged pixels in different image are identified according to the segmentation of 3D histogram. The theoretical analysis and experiment results confirm the effectiveness of the proposed method.

Key words: change detection, 3D histogram, clustering characteristic

中图分类号: