摘要
对点目标的图像变化检测,现有的变化检测技术结果往往存在着虚警过大的问题。通过深入分析多个传统的变化检测方法的特点,利用各方法的互补性,提出了利用Laplacian Eigenmap对多个方法检测结果进行降维分类的优化技术。首先把各个方法对某个像素的检测结果用向量的形式进行表示,然后利用Laplacian Eigenmap对整个图像的数据流形在低维空间展开,最后利用模糊分类进行分类。该技术有两个优势:(1)在保证现有较高检测率的同时,大大降低了结果的虚警率;(2)它极大地降低了在传统方法中由于人为阈值取舍带来的偏差风险。但该技术的不足之处是增加了计算量。
According to the high false alarm rate in the image ehange detection for point targets,an optimization method based on Laplaeian Eigenmap is proposed in this paper.We firstly express all the results of one pixel in the image by many ICD methods as a vector,and then spread the manifold which is formed by such vectors in the high dimensional space into the low dimensional space by Laplacian Eigenmap.At last these data are classified into two classes by the Gustafson Kessel,the changed points and those not.Its advantage lies in two aspects.First,it can reduce the false alarm apparently while keeps the detection rate in a high level.Second,it can also decrease the uncertainty of the result due to the unreliahle decision of the threshold value.However,such optimization increases the computational complexity.
出处
《计算机工程与应用》
CSCD
北大核心
2007年第32期196-200,共5页
Computer Engineering and Applications
基金
国家自然科学基金(No.60673090)。~~
关键词
图像变化检测
虚警优化
Laplacian特征映射
降维
image change detection
optimization of false alarm rate
Laplacian Eigenmap
dimensionality reduction
