一种基于分布式融合的多模型图像跟踪系统  

A Multi-Mode Image Tracking System Based on Distributed Fusion

作　　者：郑林[1] 韩崇昭[1] 左东广[1] 康欣[1] 

机构地区：[1]西安交通大学电信学院综合自动化所,陕西西安710049

出　　处：《系统工程与电子技术》2002年第12期1-4,共4页Systems Engineering and Electronics

基　　金：国家"973"基金项目资助课题 (2 0 0 1CB3 0 940 4)

摘　　要：设计了一种基于分布式融合的多模型图像跟踪系统。由于图像中噪声的存在、遮挡现象的发生以及被跟踪对象的变形会对某些跟踪算法产生影响 ,因此 ,一个比较理想的跟踪系统应该能够融合这些跟踪算法的优点 ,从而稳定地跟踪目标。本系统采用了 3种跟踪模式 :基于区域的跟踪模式 ;基于B样条主动轮廓的跟踪模式和基于特征点的跟踪模式。利用这 3种模式分别进行跟踪 ,并由基于区域的跟踪模式判定遮挡 ,从而决定将哪些模式的跟踪结果送到融合中心。仿真实验表明 ,这种算法能有效解决遮挡问题 ,并具有较强的鲁棒性。In this paper, a multi mode image tracking system is designed, which adopts a distributed structure to fuse the tracking results of all modes. Because of the noise in image, the presence of occlusion and the deformation of tracked target, some tracking algorithms may fail. So it is the main idea of this paper to merge the advantages of all these tracking algorithms, and then get a robust behavior. Three tracking modes are adopted in this system, i.e. region based mode, B snake based mode, and distinguished pointmode to perform tracking respectively, and the occlusion may be decided by the result of region based tracking mode, which determines which results will be sent to fusion center. The experimental results show that this system could overcome occlusion problems in tracking and has a robust behavior.

关 键 词：分布式融合 多模型 图像跟踪系统 图像处理 区域 主动轮廓 特征点 

分 类 号：TN911.73[电子电信—通信与信息系统]