一种基于帧间相关的红外弱小目标运动轨迹检测方法

针对传统帧间相关法对部分虚警轨迹难以排除的问题,研究了一种简单的帧间相关去噪算法,并提出在此基础上再次运用数学形态学方法来进一步排除虚假轨迹。算法采用形态学top-hat变换和二值化方法预处理原始图像,然后根据目标帧间相关而噪声点不相关的特点去噪,之后相加得到初步运动轨迹图像;最后采用形态学开运算,排除虚假轨迹,得到最终的检测结果。结果中检测出的目标轨迹清晰,虚警少。实验表明,该方法能有效地检测出弱小目标的运动轨迹。

New method for detecting and tracking small moving target based on target frame-correlation

To resolve the problem that some false alarm tracks are difficult to remove by using the traditional frame-correlation method,an algorithm was designed which used the mathematical morphology method to remove the false alarm tracks based on the obtained target track images in an simple frame-correlation method. First, the morphology top-hat and binaryzation method was adopted to preprocess the original image, and remove noise based on the correlation of target between frames. Then these images were added to obtain rough target tracks. Finally, the mathematical morphology for the opening operation was used to remove false alarms track and detect target trajectory. The experimental results show that this method is effective in detecting small targets.