摘要
针对跟踪算法目标遮挡后易出现跟踪漂移的问题,提出了一种跟踪学习检测(TLD)算法与Kalman滤波相结合的手势跟踪方法。在跟踪器跟踪成功后,加入识别窗的方法进行遮挡判定。产生遮挡后目标模型不再更新,学习器不再更新集合分类器。若是部分遮挡,则由TLD学习器处理;若是严重遮挡,则改由Kalman滤波算法预测目标的运动轨迹。该方法在保留TLD算法长期稳定跟踪、适应摄像机快速运动与复杂背景等优点的基础上,改善了目标遮挡后易出现跟踪漂移的问题。实验表明:提出的改进TLD算法比其他常见跟踪方法具有更加优异的性能。
To solve tracking drift problem of tracking algorithm, a gesture tracking method is proposed by combining tracking-learning-detection (TLD) algorithm with Kalman filtering. After tracker tracks successfully, blocking out decision is utilized through adding identification window. Target model is no longer updated, when generating blocking out, and ensemble classifiers is not updated by learner. Processed by TLD learner during partial blocking out ;target trajectory is predicted by Kalman filtering during severe blocking out. This method not only retains long-term stability of TLD tracking algorithm, and adapt to fast camera motion, but also overcomes tracking drift problem of target blocking out. Experimental results show that the proposed improved TLD algorithm has more excellent performance than other tracking methods.
出处
《传感器与微系统》
CSCD
北大核心
2014年第12期130-133,共4页
Transducer and Microsystem Technologies
关键词
手势
跟踪算法
跟踪学习检测
KALMAN
gesture
tracking algorithm
tracking learning detection( TLD )
Kalman