融合CamShift的TLD算法实现人脸跟踪

针对跟踪-学习-检测(Tracking-Learning-Detection,TLD)算法跟踪模块所用金字塔光流法计算量大,跟踪人脸实时性差的问题,提出融合连续自适应均值漂移(Continuously Adaptive Mean Shift,CamShift)的TLD算法提高人脸跟踪效率.改进的TLD算法框架中跟踪模块选用CamShift算法实现目标人脸跟踪,检测模块采用滑动窗法扫描搜索,再使用分类器判断目标是否存在,学习模块根据跟踪模块和检测模块的结果对比评估错误和误差,更新目标模型.将改进的TLD算法分别与CamShift算法和TLD算法进行对比试验,结果表明,融合CamShift的TLD算法实现人脸跟踪效率和准确率均高于原始两种算法,且满足实时性要求.

Face Tracking Using TLD Algorithm with CamShift Fusion

Aiming at the problem that the pyramid Lucas-Kanade optical flow method used by the TLD algorithm tracking module is computationally complex and does not meet the real-time requirements,a TLD algorithm combining CamShift is proposed to improve the face tracking efficiency.In the improved TLD algorithm framework,the tracking module selects the CamShift algorithm to achieve the target face tracking.The detection module uses the sliding window method to scan and search,and then uses the classifier to judge whether the target exists.The learning module evaluates errors according to the results of the tracking module and the detection module,and then update the target model.The improved TLD algorithm is compared with CamShift algorithm and TLD algorithm respectively.The results show that the face tracking efficiency and accuracy of the TLD algorithm combined with CamShift are higher than the original two algorithms,and the real-time requirements are met.