目标跟踪中的MS和MC自适应选择策略

针对mean shift(MS)算法不能解决非线性目标跟踪及Monte Carlo(MC)算法实时性差的问题,提出了一种自适应选择MS算法与MC算法的目标跟踪策略。首先,对目标、背景分别采样,用"对数似然法"评价每个目标区的特征对背景的可区分能力,选择区分能力强的特征作为目标。然后,引入一跟踪方式选择标志,通过计算当前跟踪窗内的目标与模板的匹配度来决定该标志的值,当目标与模板的匹配度大于某个域值时,选择实时性好的基于梯度最速下降的MS跟踪策略,以实现跟踪的实时性;否则选择基于随机采样、对目标模型没有限制的MC跟踪策略,使得位置预测结果更加准确。实验结果表明:与MC相比,本文算法在跟踪性能不受影响的前提下,有效节省了系统时间,当目标简单运动时,对于100×56像素的目标,平均计算时间由原来的82ms降低为小于1ms;与MS算法相比,该算法在牺牲一些系统时间的基础上能够更加鲁棒地解决非线性目标跟踪问题。

Adaptive choice of MS and MC for target tracking

Mean Shift deals badly with non-linear problems and Monte Carlo method increases the computational load greatly.In order to solve these problems,a new target tracking method that chooses Mean Shift and Monte Carlo method to track object adaptively is proposed.First,we sample from target region and background separately.Log likelihood method is employed to appraise the discrimination between the target and the background.Those features which discriminate target and background well are chosen as the model.Sec...