摘要
针对传统的特征点在目标追踪算法中鲁棒性不强和基于先验知识的目标追踪算法模型漂移的问题,提出一种基于光照不变特征的无模式跟踪算法。采用无模式追踪的思想,利用敏感直方图提取二值化的光照不变特征,用双向光流跟踪和全局匹配算法进行筛选得到稳定特征点,根据特征点的投票结果进行聚类确定目标中心,实现目标追踪。通过实验验证了该方法在光照变换、部分遮挡、运动模糊等情况下都表现出较好的追踪效果和鲁棒性。
Aiming at the problem that the traditional feature points are robust in the object tracking algorithm and the model of the object tracking algorithm based on prior knowledge is drifting,a model-free tracking algorithm based on illumination invariant features was proposed.Model-free tracking was used,based on locality sensitive histogram,binary description illumination invariant features were extracted.The two-dimensional optical flow tracking and global matching algorithm were used to filter the stable feature points.According to the features of the voting results,the target center was identified by cluster.Experimental results show that the proposed algorithm outperforms the other state-of-art-algorithms under illumination changes,partial occlusion,motion blur.
作者
姜可孟
曾聪文
江泽涛
JIANG Ke-meng;ZENG Cong-wen;JIANG Ze-tao(Key Laboratory of Image and Graphic Intelligent Processing of Higher Education in Guangxi,Guilin University of Electronic Technology,Guilin 541004,China;Guangxi Key Laboratory of Trusted Software,Guilin University of Electronic Technology,Guilin 541004,China)
出处
《计算机工程与设计》
北大核心
2019年第1期161-166,共6页
Computer Engineering and Design
基金
国家自然科学基金项目(61572147)
广西科技计划基金项目(AC16380108)
桂林电子科技大学图像图形智能处理重点实验基金项目(GIIP201501)
桂林电子科技大学校级教改重点项目支持基金项目(JGA201506)
广西云计算与大数据协同创新中心基金项目(YD16304)
广西高校图像图形智能处理重点实验室基金项目(GIIP201405)
关键词
目标跟踪
光照不变特征
局部敏感直方图
无模式跟踪
聚类
object tracking
illumination invariant features
locality sensitive histograms
model-free tracking
cluster
