基于自适应表面模型的概率视频跟踪算法

Probabilistic visual tracking based on adaptive appearance model

提出一种鲁棒自适应表面模型,该模型中每个像素值的变化过程由一混合高斯分布描述.为了适应目标表面的变化,这些高斯参数在跟踪期间通过在线的EM算法自适应更新;在估计目标状态时,采用了粒子滤波算法,设计了基于自适应表面模型的观测模型;在处理遮挡时,采用了一种鲁棒估计技术.多组试验结果表明,该算法对光照变化、姿态变化、部分或完全遮挡下的跟踪具有较强的鲁棒性.

A robust and adaptive appearance model is proposed, in which the value of each pixel over time is modeled by a mixture of Gaussians. To adapt to changes in object appearance, an online expectation maximization （EM） algorithm is developed to update the Gaussian parameters. When estimating the target state, particle filter is adopted, and the observation model is designed based on the adaptive appearance model. Occlusion is handled using a robust estimation technique. Numerous experimental results show that the proposed algorithm can track targets well under illumination changes, large pose variations, and partial or full occlusions.