摘要
Mean shift算法是一种重要的目标跟踪方法.在充分研究Mean shift算法的基础上,提出一种基于颜色纹理直方图的改进Mean shift跟踪算法.该方法首先计算目标图像区域中基于局部二值模式(local binary pattern,LBP)的主要纹理特征,通过提取主要特征消除背景和噪声等因素的干扰.另一方面,联合颜色和信息建立目标表示模型,可以为目标建模提供更丰富的纹理信息,目标表示更为准确,目标特征更明显区别于目标附近邻域内的背景特征.通过实验表明,改进的跟踪算法能有效提高目标跟踪精度,因此该目标跟踪具有较好的鲁棒性.
Mean shift is a useful tracking algorithm.A new tracking method based on Mean shift with color-texture histogram has been proposed on the basis of fully studying Mean shift algorithm.The improved method firstly extracts only the pixels corresponding to the main LBP features,which can reduce greatly the interferences from the smooth background and the noise.Then the color and texture features of these features to model the target,which can express more texture information for target model.And the color and texture features make the characteristics of target more obvious difference from the characteristics of background in the neighboring domain.The experimental results show that the improved tracking algorithm can effectively improve the accuracy of target tracking and make the target tracking algorithm has a better robustness.
出处
《西南师范大学学报（自然科学版）》
CAS
CSCD
北大核心
2013年第8期141-146,共6页
Journal of Southwest China Normal University(Natural Science Edition)
作者简介
梁智学（1980-），男，河南南阳人，讲师，硕士，主要从事软件工程，J2EE的研究工作．
