基于自适应跟踪窗尺度的人脸探测  

Face Detection Based on Adaptive Tracking Window Scale

作　　者：雷震[1] 王青海[1] 吴玲达[2] 薛廷梅[1] 

机构地区：[1]装甲兵工程学院信息工程系,北京100072 [2]国防科技大学多媒体实验室,长沙410073

出　　处：《计算机科学》2010年第4期281-284,共4页Computer Science

基　　金：军内科研项目基金;装甲兵工程学院创新基金资助

摘　　要：作为一种有效的迭代算法,Mean-shift具有良好的特性,在目标跟踪、图像平滑和其他计算机视觉领域得到了广泛应用。鉴于标准Mean-shift算法缺乏尺度自适应机制,而Camshift算法每次探测前需要人工选定人脸区域样本才能进行准确的探测,提出了一种用于视频中人脸探测的自适应跟踪窗算法。该算法在跟踪框内采用光照补偿和肤色分割来校正跟踪窗尺度和位置。实验表明,该算法不但具有良好的实时性,而且能较好地减少传统算法中的定位误差,更加准确地探测出视频中的人脸。As an effective iterative algorithm, Mean-shift is widely used in object tracking, image smoothing and other computer vision areas. In view of the fact that the standard Mean-shift tracking algorithm is lack of scale adaptation mechanism and Camshift algorithm needs choosing face local sample manually to track accurately before detecting, an adaptive tracking window scale algorithm was proposed for detecting face in video. The algorithm adopts illumination compensation and skin color segmentation to rectify the size and location of tacking window. The results of experiments show that the proposed algorithm not only has good real-time capability, but also reduces the location error of the traditional algorithm and achieves more accurate face detection in video.

关 键 词：人脸探测 目标跟踪 自适应尺度 MEAN-SHIFT 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]