一种综合的实时图像跟踪算法

Image-based synthesis real-time tracking algorithm

本文基于ADSP-BF561设计了一种综合的图像跟踪方法,有效地实现了低对比度下空中目标的实时跟踪。基于动态阈值的自适应双波门方法可以较好地去除云雾背景的干扰,有效地提取出空中弱小目标;基于自适应波门的最大内间方差法(OSTU)可以较好地解决大目标时自适应双波门分割目标的不完整,并且采用遗传算法(GA)对OSTU进行优化,可以较好地实现大目标跟踪的实时性。仿真实验结果表明,这种组合了两种跟踪方法优点的综合性方法,具备简单快速有效的特征,可以实现低对比度下小目标及近距离大目标的实时与稳定的跟踪。

The paper proposes an integrative and real-time target tracking scheme aiming at low contrast target in the air. A self-adaptive dual-window method based on the dynamic threshold can effectively eliminate the interference of cloud and fog background, and track and extract the small target in the air. Maximum variance between two classes （OSTU） based on self-adaptive window method can favorably solve the incomplete problem for segmentation target in big targets cases, and the real-time tracking of big target can be realized through optimizing OSTU by genetic algorithm. Simulation results show that the integrative scheme taking full advantage of the above two tracking methods is simple, rapid and effective in target tracking, and is especially capable of steady and real-time tracing for small target under low contrast condition and big target in short range.