摘要
为了减少视频目标跟踪中的累积误差,提出一种基于改进提升模型的视频目标跟踪算法。该算法结合样本有标签数据和无标签数据信息,基于半监督学习的思想,对有标签数据和无标签数据分别设计基于改进提升学习模型的分类器;将两个分类器进行加权组合,形成一个强分类器;将样本采集融合于目标跟踪的分类器学习中,有效解决了跟踪中随目标外观变化而造成的误差累积问题,提高了目标跟踪的鲁棒性。
In order to reduce the cumulative error in video target tracking, a video object tracking algorithm based on modified boosting model is proposed. A classifier based on modified boosting learning model was designed respectively by combining the labeled data and unlabeled data from the idea of semi-supervised learning. Then the two classifiers are combined to form a strong classifier. Finally, the sample collection was integrated into the target tracking classifier learning, which can effectively solve the problem of error accumulation Caused by the change of the appearance of the target and improve the robustness of target tracking.
出处
《计算机应用与软件》
北大核心
2018年第1期261-263,311,共4页
Computer Applications and Software
关键词
目标跟踪
提升模型
半监督学习
误差累积
Object tracking Boosting model Semi-supervised learning Error accumulation
