采用DBT的漂移扫描星图小目标检测方法

Small target detection method in drift-scanning image based on DBT

针对低信杂比漂移扫描星图中小目标检测易受恒星星像干扰的问题,提出一种基于单帧候选检测及多帧跟踪判决的实时处理方法。分析星图难以多帧匹配去恒星星像的原因,采用单帧处理的方式,通过构造星像端点掩膜滤除单帧检测中的虚假目标。将单帧得到的候选目标进行粒子滤波的粒子生成,运用目标运动信息构造似然函数分布图更新粒子权重,通过粒子滤波实现了漂移扫描星图中的小目标检测。实际低信杂比星图的实验结果和与相关方法的比较表明,所提方法有效提高了弱小目标的检测能力。

Aim at problem of stars have interfered with the detection of small target in the drift-scanning image of low signal-to-clutter ratio, a real-time processing method was presented, based on detecting candidate targets in single frame and tracking to judge the real target with multi-frames. The reason of filtering out the star hardly in the image through multi-frames match was analyzed, and then a single frame processing method was adopted. With the order statistics filter, the star endpoint shelter was made to filter out false targets detected in single frame. The candidate targets from the single frame processing were used for particle generation in particle filter, and the likelihood function map based on the target moving information was used for updating the particle weight in particle filter. The above operation made the particle filter can be applied to the processing of small target detection on drift-scanning images. Experiments on the actual low signal-to-clutter ratio drift-scanning image and analysis on the different related methods have shown that the method proposed has improved the ability of dim target detection.