融合粒子滤波和在线adaboost分类器的目标跟踪方法研究

传统的粒子滤波由于粒子数目有限,且模板会发生变化,模板更新时会出现漂移甚至导致跟踪失败。文章提出了融合粒子滤波和在线adaboost分类器的目标跟踪方法。将粒子滤波的跟踪结果作为正样本更新样本集,再通过样本集训练分类器检测到目标位置。然后将检测和跟踪中具有较高置信度的结果作为最终目标位置。实验证明,这种方式可以很好地解决目标重现和漂移问题。

Due to the limited number of particles and the template is likely to change,the traditional particle filter can appear drift even result in an unsuccessful tracking when the template is updated. Incorporating online Adaboost classifier with particle filter frame is proposed in this paper for object tracking.In the process of particle filter ,firstly tracking results are regarded as positive samples to update sample set, then classifiers are trained by the sample set to detect the object position.Finally the higher confidence of the detection and tracking results is exceptded to be the final object position.Experiments show that this approach can solve the problem of the object reappear and drift.