基于帧间差分法和AdaBoost算法的人脸跟踪

Face Tracking Based on Frame Difference and AdaBoost Algorithm

对于人脸在复杂背景下的检测与跟踪问题,采用帧间差分法和形态学分割出运动区域,再使用AdaBoost级联分类器对运动区域扫描,以精确定位人脸。人脸检测成功后,通过保留和及时更新地成功检测到人脸的区域,使得当人脸运动较慢时候,或人脸静止时仍然能检测跟踪成功,利用OpenCV开源库和Visual Studio 2005进行仿真,实验结果表明,在复杂的背景下,无论在人体处于移动正常、缓慢或静止的情况下,该方法都能实现单人脸检测跟踪,且跟踪速度较快。

For the face detection and tracking in complex background,proposes a novel face tracking method combined frame difference and AdaBoost algorithm frame difference are used to detect moving target regions,Then these regions are scanned by cascade classifier based on AdaBoost for more accurate face detection.When it is successfully detected,it will store and update the history face area,so that when face is moving slowly or stops moving,it also can detect the face.Carries out some simulation tests by development tool Visual 2005 and OpenCV source software library,the experimental results show that in situation with the complex background of the slowly moving face or stopping moving face,this method can successfully track the face with the good track rate.