基于二维频域投影的弱小运动目标检测方法

Dim moving object detection based on projection into the 2D frequency domain

针对光学成像弱小运动目标难以准确快速检测问题,提出了基于图像序列行列最大投影的二维频域运动小目标检测算法。首先,为了减少计算复杂度和使目标背景分离,将经过背景全局运动补偿后的视频向二维频域投影并去除频率为零的成分,从而获得含弱小运动目标和噪声的图像序列;其次,通过行列序列最大投影,得到信噪比提高的含弱小运动目标的行列图像;然后,通过主运动滤波和图像重构,检测到滤除配准误差和噪声的弱小运动目标;最后,针对强噪声干扰下的弱小运动目标检测问题进行了仿真实验,仿真结果表明,所提算法不仅可有效的检测出弱小运动目标,而且相对单一的行列分解图像在二维频域内有更高的检测信噪比。

For the dim optical object difficult to fast and accurate detection, a novel algorithm based on the row and column maximum projection of image sequences into the 2D frequency domain was proposed. Firstly, in order to reduce the computation complexity and separate the moving object from the background, through projecting into the 2D frequency domain and removing the zero frequency components, the input video with global background motion compensation was transformed into image sequences comprising of dim small moving object and noise. Secondly, after the row and column maximum projection, the higher signal-to-noise ratio image sequences with dim moving target was obtained. Thirdly, through the principal motion filtering and image reconstruction, the dim moving object was detected. Finally, the proposed algorithm was applied to the dim moving object detection in the strong background noise. The simulation experiment results show the proposed algorithm can not only effectively detect the dim moving object but also have good signal-to-noise ratio of detection.