A wide-field surveillance system with a long exposure time has a stronger capability of space target detection. However, it also produces some complicated situations that make it difficult to detect space targets; some stars appear as streak-like sources, countless object points, and possible discontinuous or nonlinear target trajectories. We present a space target detection method with high detection probability and low computational cost to overcome these obstacles. Firstly, the improved adaptive threshold method and the omnidirectional morphological filtering method are implemented to remove stars and noise. Secondly, the relative inter frame motion distance can be used as the basis for predicting the valid state transition region in each image. Finally, a state transition multistage hypothesis testing method is proposed to detect targets with linear, nonlinear, continuous or discontinuous trajectories. As demonstrated by the experimental results in simulated image sequences and real image sequences, the proposed algorithm can effectively detect space targets in wide-field surveillance with long exposure time, and has a high detection probability and low computational cost.

具有长曝光时间的宽视野监视系统具有更强的空间目标检测能力。但是，这也会产生一些复杂的情况，使检测目标变得困难。一些恒星表现为条纹状源，无数目标点以及可能的不连续或非线性目标轨迹。我们提出了一种具有高检测概率和低计算成本的空间目标检测方法来克服这些障碍。首先，采用改进的自适应阈值法和全向形态学滤波法去除恒星和噪声。其次，相对帧间运动距离可以用作预测每个图像中的有效状态转变区域的基础。最后，提出了一种状态转移多阶段假设检验方法来检测具有线性，非线性，连续或不连续轨迹的目标。实验结果表明，该算法能有效地检测出广域监视中的空间目标，且具有较长的曝光时间，具有较高的检测概率和较低的计算量。