摘要
目标跟踪技术是视频检测技术中一个十分重要的组成部分,为此,提出一种基于特征点的快速跟踪算法。该方法避免了困难的目标分割过程。采用两次帧差共同确定角点选择区域,利用Moravec算法提取合适角点;采用一种特别设计的包含不平滑区域的结构化模板获取更好的匹配点;利用预测点缩小搜索范围,降低计算复杂度和时间复杂度。实验证明该算法能够快速实现目标的实时跟踪,跟踪准确度高,对不同的场景都具有良好的鲁棒性。
Target tracking technique is a very important part of video detection technology, for this reason, a fast tracking algorithm is proposed based on feature points without difficult image segmentation procedure. Two images gotten by frame difference are used to identify areas that are suited to extract corner, and feature points are extracted using an improved Moravec algorithm. A specially designed template with no-smooth areas from the comer neigh- borhood is used to track the feature points through the image sequences. It can get better matching point. Forecasting matching points are used to de- crease searching range, reduce the computational complexity and time complexity. Experiments show that this algorithm can quickly achieve the real-time tracking to targets. The tracking results have high accuracy and good robustness to different conditions.
出处
《电视技术》
北大核心
2014年第5期34-37,共4页
Video Engineering
基金
陕西省科技攻关计划项目(2011k06-32)
中央高校基本科研业务费专项(CHD2011TD012)
长安大学科技创新重点项目
陕西省道路交通智能检测与装备工程技术研究中心开放基金项目
关键词
目标跟踪
角点提取
块匹配
模板更新
target tracking
corner extracting
block matching
template updating
