一种快速人脸跟踪算法的研究

A New Algorithm of Quick Face Tracking Based on Motion Estimation

提出了一种新的在连续运动场景中人脸的检测和跟踪方法.首先采取FloatBoost算法检测人脸,以提高检测速度和精度,然后运用运动学原理及运动估计的思想,利用时间序列分析中移动平均法和指数平滑法预测下一帧图像中跟踪目标的运动位置区域,以减少图像搜索区域,降低处理资源的消耗,达到实时跟踪的效果.仿真实验中,利用MATLAB进行人脸的检测、跟踪实验,并运用本文算法与FSA,CPME算法对跟踪目标物体的时间进行了对比实验.实验结果表明,本文所提出的方法具有良好的实时性和准确性.

A new method for face detection and tracking in sequential moving state was proposed.The FloatBoost algorithm was used for face detecting to enhance its speed and accuracy.Then,based on the theory of kinematics and motion estimation,the motion averaging method and exponential smoothing method in time-series analysis were used to predict the moving location area of tracking object in next frame image so as to reduce the searching region and decrease the consumption of processing resource,thus implementing the real-time tracking effect.In the face detecting/tracking simulation by MATLAB the algorithm proposed was compared with the FSA and CPME algorithms in respect to the time required for tracking an object,and the results showed that the former is superior to the latter two in real-time response and accuracy.