基于加权Lucas-Kanade算法的目标跟踪

Object Tracking Based on the Weighted Lucas-Kanade Algorithm

针对传统的序列图像目标跟踪方法难以适应复杂背景干扰、目标形状变化以及目标位置非规则抖动的问题,提出了一种基于加权Lucas-Kanade算法的目标跟踪新方法。首先引入搜索模板,估计出目标在实时图像中的位置并将其作为加权Lucas-Kanade算法的迭代初始值,然后计算权值函数,利用当前模板和初始模板进行两次跟踪,得到目标的准确位置。最后实现了在目标形状及背景变化下的三种模板更新。大量实测数据的实验结果表明,本文所提的方法有效地实现了对地面复杂场景中形变目标的稳定跟踪。

The traditional object tracking methods are difficult to deal with the interference of complex background,the variety of object shape and the irregular change of object position.A new object tracking method based on the weighted Lucas-Kanade algorithm is proposed.Firstly,a reasonable estimation of object position in current frame is obtained according to the search template.This position is used as initial iteration parameter of the weighted Lucas-Kanade algorithm.Secondly,the weights function is calculated and the accurate object position is achieved by tracking the current template and the initial template in current frame.Finally,the template update strategies under the complex background and the object shape variety are implemented.Experimental results based on a large amount of measured data show that the proposed method can effectively realize stable tracking of object in complex background.