基于决策融合和模型更新的红外目标跟踪方法

Infrared Visual Tracking Via Decision Fusion and Model Updating

红外图像中的目标跟踪技术在导弹制导等领域有着广泛的应用,是现代化军事实力不可缺少的一部分。传统方法主要构建目标的轮廓纹理模型,但对于模糊目标、快速移动目标或存在外界遮挡等情况并不鲁棒。提出一种基于动态决策融合和自适应模型更新机制实现的跟踪算法,利用轮廓纹理和红外能量分布信息共同描述目标,在不同情况下动态计算各支路权重以及模型的更新因子,从而充分利用了不同模型在各种情况下的优势,并避免错误的目标表观信息更新至模型。实验表明,提出的方法有效提高了算法在目标快速移动、形态模糊和外界遮挡等情况下的鲁棒性。

Infrared visual tracking has a wide range of applications in many fields such as missile guidance. It is one of the most important parts of modern military strength. Traditional tracking approaches mainly construct the contour and texture models of the target, which is not robust enough to fuzzy target, fast-moving target, occlusion and other situations. This paper proposes a tracking algorithm based on the dynamic decision fusion strategy and the adaptive model updating scheme. The contour texture and infrared energy distribution information were used to describe the target together. The weights of each branch and the update factor of the model were dynamically calculated under different conditions so that different models were fully utilized under different situations. In addition, the proposed method could prevent the model from the wrong target apparent information. Experimental results demonstrate that the proposed approach is significantly robust to fast-moving, occlusion and fuzzy targets.