CAJ | 학술논문

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。
침대광학성상약소운동목표난이준학쾌속검측문제，제출료기우도상서렬행렬최대투영적이유빈역운동소목표검측산법。수선，위료감소계산복잡도화사목표배경분리，장경과배경전국운동보상후적시빈향이유빈역투영병거제빈솔위령적성분，종이획득함약소운동목표화조성적도상서렬；기차，통과행렬서렬최대투영，득도신조비제고적함약소운동목표적행렬도상；연후，통과주운동려파화도상중구，검측도려제배준오차화조성적약소운동목표；최후，침대강조성간우하적약소운동목표검측문제진행료방진실험，방진결과표명，소제산법불부가유효적검측출약소운동목표，이차상대단일적행렬분해도상재이유빈역내유경고적검측신조비。
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.