中文摘要：

在线运动目标跟踪是目前模式识别领域的一个难点问题，目标物体角度、姿态、远近距离变化以及遮挡等给鲁棒在线跟踪算法提出了苛刻的要求，单一算法很难有效处理所有问题．多方法集成是实现鲁棒在线跟踪的一种有效手段，为此提出了一个集成on-line boosting、基于归一化互相关的模板匹配法和粒子群优化算法的自适应目标跟踪算法框架．其中，on-line boosting是基本的跟踪算法；基于归一化互相关的模板匹配法有效避免了on-line boosting过多的错误更新；而基于粒子群优化算法的跟踪策略提高了系统对快速运动、外观变化的适应能力，同时也为模板的更新提供了保障，三种算法形成了有效互补，在稳定性和可塑性之间达到了一种平衡．在不同视频测试序列上的实验结果表明，该算法有效地缓解了自适应性和“漂移”之间的矛盾，能够实时地完成复杂的跟踪任务．

英文摘要：

On-line object tracking has been a difficult problem in the literature of pattern recognmon, clue to harsh requirements for robust on-line tracking algorithm which can handle changes of viewpoint, pose variation, distance and occlusion. A single approach can＇t handle all the problems, thus an integration of multiple methods is an effective strategy for robust on-line tracking. An adaptive object traeking framework was proposed which integrated on-line boosting with normalized cross-correlation based template matching and particle swarm optimization. Among these three methods, on-line boosting was thebasic tracking algorithm; Template matching was employed to effectively prevent on-line boosting from making too many wrong updates, while particle swarm optimization based tracking strategy improved the adaptability to rapid movements, fast appearance variations and meanwhile guaranteed the update of the templates. The algorithms were complementary and kept a balance between stability and plasticity. The experimental results of different test sequences show that dilemma between adaptability and drifting and successfully the proposed approach efficiently alleviates the tracks objects in real-time.