基于主从记忆空间模型的时空上下文跟踪算法

SpatioTemporal Context Tracking Algorithm Based on Master-Slave Memory Space Model

提出了一种基于主、从记忆空间模型的时空上下文跟踪算法。该算法将人脑记忆机制融入STC算法的时空上下文模板更新过程,通过构建主、从记忆空间,形成基于记忆的模板更新策略。同时,通过计算置信图多峰值点求取目标位置,提高目标跟踪精度。实验结果表明,所提出的算法可解决目标被遮挡、姿态突变、短暂消失后重现等条件下的跟踪精度下降问题,有利于实现鲁棒性、高精度运动目标跟踪。

A spatio-temporal context tracking algorithm based on master-slave memory space model was proposed in this paper.The algorithm introduced the memory mechanism into the template updating process of STC algorithm,and constructed two memory spaces for the master and the slave.Finally,a memory based template updating strategy was formed.Meanwhile,this algorithm determined the target location based on confidence map multi candidate,it improved the tracking accuracy.Experimental results show that the algorithm can effectively solve the problem of abrupt change of target pose,recurrence,occlusion and so on,which is help to achieve high-accuracy and robust moving object tracking.