自适应模型更新的多特征融合目标跟踪算法

Object Tracking Algorithm Based on Adaptive Observation Model Updating and Features Fusion

针对目标在运动过程中存在遮挡、光照变化、背景因素等复杂情况下的跟踪问题,提出了一种多特征融合的跟踪算法;利用背景加权后的联合直方图来描述目标的灰度和纹理特征信息,提出一种多帧加权组合的模板更新策略,对模板特征分布进行自适应更新,基于当前粒子特征信息可信度加权设计了一种自适应特征融合观测模型,并结合到粒子滤波算法中,从而提高了跟踪算法的场景适应能力;实验结果表明:与基于单一特征的算法相比,该算法的适应性更强,能有效跟踪复杂场景下的运动目标。

When the moving object is occluded, or illumination changed, or there are background disturbances, it is hard to track the moving object. A features fusion method is proposed to solve these problems. Used united histogram to describe the grayseale and vein information of the object, presented a multiple picture model updating, designed a self--adaptive features fusion observational model based on the credible probability weighted of optimal particle features and combine it with particle filter to improve the scene adaptability of the method. The experimental result shows that the availability of this method is superior to those methods based on single feature, and can track moving object effectively in complex scene.