基于局部稀疏的目标跟踪方法

Object tracking based on local sparse representation

由于大多数稀疏表示的目标跟踪方法仅考虑目标的全局表示,容易导致目标丢失,因此对局部稀疏方法进行研究,提出一种基于局部稀疏表示的追踪方法。通过对目标区域局部图像块进行稀疏编码来表示目标观测模型,这种基于局部特征的跟踪器能够应对目标的外观变化;考虑遮挡因素的影响,对有遮挡的图像块做特殊处理;采用逻辑回归分类器进行分类,区分背景和目标物体,提高目标跟踪的准确度。对各种视频图像序列进行测试,测试结果表明,该方法具有更好的性能。

most of object tracking methods based on sparse representation only consider target＇s global representation, leading to the tracking failing, so by studying the local sparse representation, an object tracking method based on it was proposed. A local sparse appearance model was proposed, which making use of the local information of the target by sparse coding. The tracker with local features was able to cope with target appearance changes. An occlusion handing scheme was taken into account. A lo- gistic regression classifier was used to distinguish between background and target object. Various challenging videos were used to test. Results demonstrate this method has better performance than many state-of-the-art algorithms.