基于人脸检测与跟踪的广告效果评估系统  

Advertisement effect evaluation system based on face detection and tracking

作　　者：邢延超[1] 强文萍[2] 

机构地区：[1]青岛理工大学通信与电子工程学院,山东青岛266033 [2]青岛大学信息工程学院,山东青岛266071

出　　处：《计算机应用》2009年第10期2700-2702,共3页journal of Computer Applications

摘　　要：公共场所各种广告对商品的宣传至关重要,但对广告效果的评价则缺乏有效的自动分析手段。针对这一问题提出一种基于计算机视觉处理的广告效果评价系统。该系统通过采集广告前方人群运动场景的视频,检测并跟踪视频中关注广告的人脸,记录其关注时间,并根据关注者数量、关注时间、关注模式分析等,对该广告的受众反馈进行量化评价。人脸检测采用AdaBoost分类器,人脸跟踪采用粒子滤波器。人脸检测与跟踪相互补充和修正,提高了人脸跟踪的可靠性。实验结果验证了该方法的有效性。Various advertisements in public are important for commodities promotion, bat it＇s still hard to automatically analyze the effect. This paper presented a digital video processing based evaluation system, which captured the scene in front of the advertisement, then performed statistics on the number of the passengers, especially the ones looking to the advertisements, recorded their watching time intervals. With the above parameters, one could estimate the effect of the advertisements. Face detection adopted AdaBoost classifiers and face tracking used particle filters. The detection and the tracking are supplementary and could improve the tracking reliability. The experiments proved the effectiveness.

关 键 词：广告效果评价 人脸检测 粒子滤波器 运动轨迹 ADABOOST算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]