关键词: 红外图像序列, 块匹配, 背景估计, 背景差法, 聚类识别, 目标检测

Abstract: According to the airborne jitter，dynamic background and internal noise of the airborne infrared weak targets，target detection has become a difficult problem.In this article，a kind of convolution template，which combined with filtering algorithm，is presented for image preprocessing.Then in order to estimate the background，the reflection model of global motion is proposed.What’s more，by using adaptive background subtraction algorithm and clustering recognition，the suspected targets can be extracted.Those suspected targets are used for pipeline regionalization，that is，suspected targets trajectory with continuity are the real goals，otherwise are false targets.

Key words: infrared image sequence, block matching, background estimation, background difference, clustering recognition, target detection