基于聚类分析的红外弱小目标检测

摘要

针对深空背景下的红外弱小目标检测，提出了一种基于聚类分析的目标检测方法，该方法将经过背景抑制的连续几帧图像构造组合帧，基于目标的运动特性，对分割后的组合帧进行聚类分析，从而检测到弱小目标并同时获得目标运动轨迹，再对检测结果进行聚类检验，从而去除虚假目标，降低虚警率。实验结果表明该算法对多目标的检测有较高的鲁棒性，且相对于传统的小目标检测算法有更高的检测率和较好的实时性。%In order to detect small targets in deep space, a detecting algorithm based on cluster analysis is proposed. The ap-proach forms a composite frame by a few of infrared images which have been preprocessed, and the composite frame is clus-tered after it is segmented based on moving targets feature. Accordingly, small targets and the trajectories of moving targets can be obtained. The false positive objects are deleted by followed validation process. Experimental results show that this method is robust in small targets detection. Compared with some traditional methods, good performance can be obtained in detection rate and detection efficiency.