基于颜色-纹理特征的目标跟踪  

Target tracking based on color and the texture feature

作　　者：姚原青 李峰[1] 周书仁[1] 

机构地区：[1]长沙理工大学计算机与通信工程学院,湖南长沙410004

出　　处：《计算机工程与科学》2014年第8期1581-1587,共7页Computer Engineering & Science

基　　金：国家自然科学基金资助项目(60973113);湖南省自然科学基金资助项目(12JJ6057);长沙市科技计划资助项目(K1203015-11);湖南省标准化战略项目(2011031)

摘　　要：针对传统的基于颜色特征目标跟踪算法在一些复杂场景中存在的跟踪不稳定性,提出一种基于颜色-纹理特征的目标跟踪算法;在传统的基于颜色Mean-shift的目标跟踪算法中加入纹理特征,在提取目标颜色特征的同时提取目标的纹理特征,并且采取串接原则,在搜索目标新位置时仍然沿用传统的基于颜色的均值漂移跟踪算法,但在每一次迭代过程搜寻目标最佳的位置点即特征相似最大的区域时,利用纹理特征来实现,并且采用八邻域搜索法(候选区域周围扩大八个大小相等的区域)来解决部分遮挡的问题。通过对比实验表明,该算法在复杂场景中表现出的实时性和鲁棒性较好。In some complex scenes of tracking,the traditional color feature based target tracking algorithm may have the unstable tracking.To solve this problem,a color and texture feature based target tracking algorithm is proposed.The texture feature is added in the traditional color Mean shift based target tracking algorithm,and both the texture and the color features are extracted together.According to the sequence principle,the traditional color based mean drift tracking algorithm is still used to search the new position of the target,but,in each iteration process,the texture feature is used to search the optimal target position which has the largest similar feature.And the eight neighborhood search method （8 areas with the same size are expanded around the candidate area） is applied to solve the problem of partial occlusion.The comparative experiments show that the algorithm has better robustness and realtime performance in complex scenes.

关 键 词：目标跟踪 颜色特征 均值漂移 纹理特征 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]