星空背景下的运动点目标轨迹提取

Trajectory extraction algorithm of moving point targets in star background

针对星空图像中运动点目标轨迹难于提取的问题,提出了有限帧最小二乘轨迹关联算法。采用相邻帧差分、自适应阈值分割、连通域处理和质心计算等步骤实现了疑似目标的定位。根据星点目标运动轨迹近似为直线、运动的连续性和相关性等特点,利用有限帧最小二乘线性拟合预测目标位置,以预测位置为中心开搜索窗,关联匹配真实目标,并通过置信度检验确认并提取目标运动轨迹。整套算法简单,易于实现,通过实拍星图实验,验证了该算法效果好,实时性强。

Based on the analysis of the characteristics of star image and point target, an investigation about the algorithm for trajectory extraction of moving point targets in sequence star images is carried out. A track-association algorithm based on the limited frames Least Squares method is put forward. In the single frame processing, after these steps of image difference of related frames, auto threshold segmentation, calculating 8-directions-connected fields, centroid method of gray-scale weighted etc are carried out in turn, doubted targets are rapidly located. In the sequence image frames processing, according to the supposition that target moving trajectory is approximate beeline, continuity and relativity of movement, target position are predicted by Least Squares method. The real targets are associated and matched in the searching windows which midpoints are located on the predicted positions. The real target moving trajectory is extracted by doubting grade checkout. The complete set of algorithms are used to process sequence star images that are focused in the real condition, and obtain satisfied results.