针对凝视观测模式下光测图像中弱小目标的检测问题, 提出一种多帧频域特征累积的目标检测方法。利用恒星匹配、临帧差分等方法消除背景变化的影响, 通过图像时频域特征分析, 推导出递归形式频域特征累积方程, 并由此检测图像中的弱小目标。实验结果表明采用多帧频域特征累积方法能将图像信噪比提高3倍以上, 有利于检测出弱小目标。

Aiming at the problem of tiny object detection in stare observation image sequence, a new method based on frequency feature accumulation was proposed. First, star matching and difference of neighbor frames were used to eliminate background of images, and then by analyzing time-frequency feature, the method for accumulating frequency feature of image was applied in tiny object detection. The experiment shows that the method proposed can increase the signal-noise ratio (SNR) of image by 3 times, and favors tiny object detection.