摘要
传统的用户行为监控方法判断误差率较大,为此,提出以人脸识别技术为基础,对公共图书馆用户行为自动监控的方法,通过计算分析图像多维空间数据,得到人脸识别所需要的核函数与密度梯度,对比面部帧和背景模型,得到面部矩形区域划分结果,凭借特征计算构建弱分类器,将多组弱分类器组合构成强分类器,再将几组强分类器组合成级联分类器,同时通过精确定位算法对用户图像内眼睛与嘴巴位置定位,完成对公共图书馆用户行为的监控。实验结果证明,所提方法能够提升行为监测结果的精准度。
The error rate judge of the traditional user behavior monitoring method is large.Therefore,based on face recognition technology,the method of automatic monitoring of user behavior in public library is proposed.The kernel function and density gradient of face recognition are obtained by computing and analyzing multi-dimensional space data.The background and foreground of public library images are distinguished by foreground detection,facial frame and background model are compared,the result of rectangular region division is obtained,rectangular region integral map is obtained by feature calculation,and weak classifier is constructed,and multiple groups of weak classifiers are combined to form strong classification.Then several groups of strong classifiers are combined into cascaded classifiers,and the position of eyes and mouth in user image is located to monitor the user behavior of public library.The experimental results show that the proposed method can improve the accuracy of behavior monitoring results.
作者
薛鑫
杨自勉
XUE Xin;YANG Zi-mian(Shaanxi College of Communication Technology,Xi'an 710018 China;Warranty Branch of Xi'an Public Transportation Group Co.,Ltd.,Xi'an 710018 China)
出处
《自动化技术与应用》
2023年第8期28-33,共6页
Techniques of Automation and Applications
关键词
人脸识别技术
公共图书馆
用户行为监控
矩形特征
运动目标检测
face recognition technology
public libraries
user behavior monitoring
rectangle features
moving object detection