基于 SIFT 特征和模板更新的粒子滤波目标跟踪算法

摘要

In traditional particle filter algorithm, the weight of each particle is updated only by the colour feature of the object, which may easily lead to error tracking when the object and the background have similar colour distribution or the object is occluded.Scale-invariant features are highly discriminative, but to use SIFT only is insufficient to describe small targets.A new method is proposed in this paper to handle with these two situations, in which the target model is built by SIFT feature and colour feature, and the particle filter is integrated to achieve object tracking.In order to avoid error updating of the target model, in this paper whether the colour target model is to be updated or not depends on the number of matching feature points between the tracking result in current frame and the SIFT target model.Experimental results show that the proposed method can effectively improve the tracking precision especially when the object is occluded or under clutter background with similar colours.%传统的粒子滤波算法利用目标的颜色特征对粒子权值进行更新，当背景与目标的颜色分布相似或者目标被遮挡时，易发生误跟踪。尺度不变特征具有较高的独特性，但是仅使用SIFT特征不足以对小目标进行描述。针对这两种情况，提出一种利用SIFT特征和颜色特征建立目标模型，结合粒子滤波实现目标跟踪的新方法。为了防止目标模板的误更新，根据当前帧跟踪结果与SIFT目标模板中特征点的匹配数目决定是否对颜色目标模板进行更新。实验结果表明，当目标被遮挡或者位于有相似颜色的杂乱背景时，提出的方法能有效提高跟踪的精确度。