基于目标模型自适应更新的mean shift跟踪算法

Mean Shift Tracking Arithmetic Based on Adaptive Update of Object Model

本文提出一种自动更新mean shift跟踪模型的算法。该算法采用Kalman滤波对跟踪系统下一帧的目标模型进行预测,通过对滤波残余误差样本的假设检验,提出一种更新机制。实验结果表明,跟踪系统可以在目标被遮挡或形状改变的情况下,有效地更新目标模型,实现实时目标跟踪。

This paper proposes a arithmetic that automatically updating mean shift tracking model. This kind of arithmetic uses the Kalman filter for estimate the object model in the next frame of the tracking system, and presents a robust criterion with the t hypothesis testing with the samples from the filtering residuals. Experimental results show that the tracking system can effectively update the object model under the circumstances of severe occlusion or appearance change, and achieve object tracking in time.