智能录播系统中站立人脸检测定位的实现  

Realization of Detection and Location of Standing Face in IntelligentRecording and Broadcasting System Computer Engineering and Applications

作　　者：刘小芳 魏伟波 谭璐 LIU Xiao-fang, WEI Wei-bo, TAN Lu(School of Computer Science and Technology, Qingdao University, Qingdao 266071, Chin)

机构地区：[1]青岛大学计算机科学技术学院,青岛266071

出　　处：《青岛大学学报（自然科学版）》2018年第1期85-91,共7页Journal of Qingdao University(Natural Science Edition)

基　　金：国家自然科学基金(批准号:61170106)资助

摘　　要：针对智能录播系统中站立检测的现状和需求,提出了一种站立人脸的检测定位算法。算法通过帧差法确定活动区域,然后对活动区域光照补偿、肤色检测初步筛选人脸区域,结合人脸特征以及人脸面积占比等,综合区域划分,实现了站立人脸的检测定位。实验结果表明,该算法具有较高的检测准确率和鲁棒性。According to the present situation and demand of standing detection in intelligent recording system,a location algorithm for standing face detection is proposed.The algorithm determines the active region by frame difference method.Then preliminary screening the face area by illumination compensation and skin color detection,combined facial features and face area ratio,integrated region partition,realized the detection and location of standing face.It is showed that the proposed algorithm has higher detection accuracy and robustness by experimental results.

关 键 词：帧差法 人脸检测 光照补偿 肤色检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]