一种快速多人脸跟踪算法

提出一个基于Mean Shift的实时多人脸跟踪算法。通过引入自适应目标跟踪窗口，改进了Mean Shift算法的目标连续跟踪性能；提出序贯跟踪法解决多人脸跟踪过程中目标发生粘连重叠的问题；引入多辅助信息解决了相邻两帧中人脸的对应问题。为进一步提高整个算法的跟踪速度和鲁棒性，引入卡尔曼滤波器对目标进行预测。实验结果表明该算法具有很好的实时性和跟踪效果。

Improved Mean Shift real-time multiple faces tracking algorithm

A real-time multiple faces tracking algorithm based on Mean Shift algorithm was presented in this paper. An adaptive kernel window was adopted for each object, and a ranked sequential tacking strategy was proposed to solve the occlusion problem. Multi-accessory information was introduced to match the tracked objects and target windows. Furthermore, in order to improve the tracking speed and robustness, Kalman filter was introduced to predict the position of the object window. The experimental results show that the proposed multiple faces tracking algorithm can track multiple faces robustly and in real-time.