基于分段加权的反向稀疏跟踪算法研究

为提高稀疏表示跟踪模型性能,提出一种分段加权的反向稀疏跟踪算法,将跟踪问题转化为在贝叶斯框架下寻找概率最高的候选对象问题,构造不同的分段权重函数来分别度量候选目标与正负模板的判别特征系数。通过池化来降低跟踪结果的不确定性干扰,选择正负模板加权系数差值最大的候选表示作为跟踪结果。实验表明,在光照变化、遮挡、快速运动、运动模糊情况下,所提出的算法可以确保跟踪结果的准确性和鲁棒性。

Research on Piecewise Weighted Inverse Sparse Tracking Algorithm

To improve the performance of sparse representation tracking model, a piecewise weighted inverse sparse tracking algorithm is proposed, which translates the tracking problem into finding the most probable candidate target within Bayesian framework. Different piecewise weighted functions are constructed to separately measure the discriminant characteristic coefficients of the candidate target with the positive and negative templates. The pooling is utilized to reduce the uncertainty of the tracking results of interference, then the candidate represented by the biggest difference between the positive and negative template weight coefficients is chosen as the tracking result. Experiments indicate that the proposed algorithm can ensure the accuracy and robustness of tracking results in case of the light changes, occlusion, fast motion, motion and blur.