摘要
针对周扫条件下的运动目标检测出现的问题,提出了多种解决办法,并从中选择最优方案。首先根据系统红外图像特点,使用投影匹配法进行图像配准;然后针对现有运动检测算法中存在的问题,结合帧间差法和经典自适应背景差分法的优点,提出一种基于分级自适应背景差分的运动检测方法,使得伪运动目标明显减少;最后针对剩下的伪运动目标,借鉴图像配准方法,提出了一种基于最小平均灰度差判定真假运动目标的方法,有效剔除了伪运动目标。实验结果表明所提出的算法能快速有效的解决周扫模式所带来的各种问题。
Aiming at the problems of infrared moving object detection under the condition of rotating scan system, several methods were presented, and an optimal solution was selected. Firstly, according to the characteristics of infrared image, a projection matching method was used for the image registration. Aiming at the faults of traditional moving object detection methods, combining the advantage of frame difference and the classic self-adaptive background difference algorithm, a new approach was presented based on differentiated self-adaptive background difference algorithm, which made the pseudo moving object decreased obviously. At last, learning from the image registration, a method based on the least subtraction of mean gray value was proposed to remove the pseudo moving object effectively. Experimental results show that the presented algorithms can resolve the problems brought by the rotating scan mode quickly and effectively.
出处
《红外与激光工程》
EI
CSCD
北大核心
2014年第3期967-972,共6页
Infrared and Laser Engineering
关键词
运动目标检测
帧间差分
背景差分
投影匹配
moving object detection
frame difference
background difference
projection matching
