空间临近目标行为辨识方法

针对侦察卫星通过变轨等方式抵近目标的对天观测行为，提出了一种基于视觉特性的临近目标行为辨识方法。通过对序列图中抵近目标上特征点进行跟踪以得到其轨迹变化；针对目标可能存在的异常行为在图像当中的表征进行分析，结合已得信息构建特征参数并对目标行为进行判断；最后通过试验对辨识算法进行验证。试验结果表明：该算法能对不同光照条件下的连续图像进行处理，实现对临近目标通过主体姿态变化进行对天侦察行为的有效辨识。

Behavior identification of space adjacent targets

Aiming at the observation behavior of reconnaissance satellite approaching the target by changing orbit, a method of behavior identification for near target based on visual characteristics was proposed. By tracking the feature points of the adjacent target in the image sequence, the trajectory change was obtained. Afterwards, the representation of the possible abnormal behavior of the target in the image was analyzed, and the characteristic parameters were constructed based on the obtained information to judge the target behavior. The experimental results show that the algorithm can process the continuous images under different illumination conditions, and realize the effective identification of the space reconnaissance behaviors with attitude changes of adjacent targets.