摘要
远距离目标检测具有重要的军事应用价值,而多传感器能提供互补和冗余信息,在远距离目标检测中具有独特的优势。针对可见光和红外热图像序列中的远距离目标,提出了一种基于信息融合的目标检测方法。该方法通过帧间差累积,在两种传感器的图像中确定了运动目标区域,定义了运动目标区域可信度度量,利用融合的规则确定了最后的运动区域。在运动区域中利用简单的分割方法将目标提取出来,最后经过特征级融合得到了检测结果。本方法不需要进行背景估计,多传感器信息的冗余增强了检测结果的可靠性,实验结果证明了本文方法的有效性。
Long-range target detection takes an important part in military applications. Multisensor can provide complementary and redundant information and has advantage in long-range target detection. In this paper, the long-range target detection based on data fusion in visual and thermal infrared image sequences is proposed. In this way, the moving targets range in two sensor image sequences are determined by using a frame difference accumulation, and the reliable measure at the moving target range is defined and the final moving range is determined with the data fusion rules. The targets at the moving range are extracted by using a simple segmentation. The detection results are acquired by feature-level image fusion. The multisensor redundant information enhances the reliability of the detection. The results show that this approach is feasible and robust.
出处
《红外技术》
CSCD
北大核心
2006年第12期695-698,共4页
Infrared Technology
基金
国家重点实验室基金资助(51483040105QT5118)
关键词
可见光图像序列
红外图像序列
帧间差
信息融合
visual image sequences
thermal infrared image sequences
frame difference
data fusion
