基于稀疏表示的红外小目标跟踪

Tracking infrared small targets based on sparse representation models

针对云天背景下红外小目标跟踪易受噪声影响,传统目标模型对噪声敏感等问题,提出一种贝叶斯推理框架下使用稀疏表示建模红外小目标的跟踪算法。该算法结合目标模板向量和正负琐碎向量构建目标稀疏表示模型,在贝叶斯推理框架下使用图像子块系数向量的目标模板重构误差作为观测模型,实现小目标的跟踪。实验证明了算法的有效性和鲁棒性。

In order to improve the ability of tracking infrared small targets under a cloudy sky background. This paper proposes a novel Bayesian inference tracking framework based on sparse representation. To track the targets at each new frame, all of target candidates are sparsely represented by targets templates and trivial templates respectively, and which candidate with the smallest projection error is taken as the tracking target. A Bayesian framework is used in the whole tracking procedure. Several experiments demonstrate the proposed algorithm’s efficiency.