【摘要】 提出一种基于Gabor小波特征而进行分类的目标跟踪方法。传统的基于Gabor特征的跟踪方法主要利用目标边界特征,但目标外观发生巨大变化时,跟踪往往会失效。在这种情况下,提出的新算法同时利用背景和前景(目标)的Gabor特征信息,将目标跟踪问题转换为前景和背景的分类问题。实验结果表明,该方法能够在视觉监控系统中有效的跟踪人,汽车等目标,并且对光照,目标大小改变具有很好的鲁棒性。更多还原

【Abstract】 Many Gabor wavelet-based approaches make use of the boundary of the object,in which ca- ses tracking may fail when great changes occur in target appearance.The paper presents a tracking method based on Gabor wavelet features and classification of targets,which combines background and foreground infor- mation,thus turning the problem of tracking into one of foreground-background classification.The experimen- tal results show that this approach is robust for targets with zooming,rotation and illumination changes.更多还原