一种视觉跟踪中的模板更新策略

针对复杂场景中的目标外观和背景变化引起的模板 更新问题, 提出了一种新的视觉跟踪模板更新策略,用以提高目标模板正确性。算法利用特征信息在 时间和空间上的区别和变化,进行特征子量分类更新,避免了模型过更新,提高了目标模型 的容错能力,使更新带来的误差尽量小,以适应目标和背景信息的不断变化,在一定程度上 提高了跟踪算法的精准度和鲁棒性。实验结果表明,本文方法在视频跟踪系统中具有优越的 性 能,可以在目标运动、变化和遮挡情况下实现鲁棒跟踪。

A template updating strategy in visual tracking

In visual tracking system,template upd ating has been a very hot and challenging topic in the visual tracking algo rithm s based on pattern matching,and has a great value in practical application and research.In order to cope with the problem of template updating due to variations of object appearance and background under complex environment,a novel template updating strategy of visual tracking is proposed in this paper to enhance the precision o f the target template. The proposed method can update the classified sub-model by the distinguishes an d changes of feature information in both time and space,and it also avoids the over-updatin g of template.It makes fewer errors during the updating,and improves the fault tolerance ability of the target template.The adaptive template can be used in the case of appearance and backgro und changes,and can improve the tracking accuracy and robustness in a certain extent.Experimental r esults show that the proposed method exhibits better performance in the visual tracking system,a nd is robust to pose,illumination,deformation and occlusion.