基于改进On-line Boosting算法的视频目标跟踪  

Video Target Tracking Based on Impoved On-line Boosting Algorithm

作　　者：蔡明琼 郭太良[1] 姚剑敏[1] 

机构地区：[1]福州大学平板显示技术国家地方联合工程实验室,福建福州350116

出　　处：《电视技术》2015年第16期65-68,共4页Video Engineering

基　　金：国家"863"重大专项(2013AA030601);福建省自然科学基金项目(2011J01347)

摘　　要：针对目前基于在线学习的On-line Boosting算法用于视频目标跟踪时对于快速移动的目标,容易引起跟踪漂移的问题,提出一种将Surf算法融合于On-line Boosting的Surf-Boosting视频目标跟踪算法。该算法在原先的On-line Boosting算法的基础上增加跟踪漂移判断,对已跟踪漂移的视频帧使用Surf算法进行目标定位,将Surf定位到的目标作为正样本放到后续On-line Boosting算法中继续跟踪学习。实验结果表明,该方法能够很好地抑制原有算法的跟踪漂移问题,在跟踪过程中的正确率达到98%,实现对快速移动目标的正确跟踪,并具有很好的鲁棒性。In view of that the present way of On-line Learning video target tracking algorithm based on the On-line Boosting algo- rithm would cause tracking drift easily, the thesis put forward a kind of new improved algorithm called Surf-Boosting which com- bine Surf algorithm to the On-line Boosting algorithm in video target tracking . Compared with the original On-line Boosting algo- rithm, the imprnved algorithm add the judgment of tracking drift and can locate target by using Surf algorithm when tracking drift. After that the target located by surf algorithm can be learned as a positive sample in follow-up tracking. The experimental result whose accuracy is 98% show that the improved robust method can restrain the tracking drift and realize good tracking on fast moving target effectively.

关 键 词：ON-LINE BOOSTING SURF 快速移动 目标跟踪 

分 类 号：TP227[自动化与计算机技术—检测技术与自动化装置]