防欺骗人脸识别考勤系统研究与设计

发布时间：2018-11-11 17:53

【摘要】：相比于其它生物识别技术,人脸识别具有难以比拟的优势,具有更广泛的应用领域,是机器视觉中的研究热点。本文将基于人脸识别技术进行考勤系统的研究与设计,并针对现有的人脸识别系统不具有防欺骗性,易受照片和视频欺骗攻击的特点,提出了改进方法。首先,采用AdaBoost人脸检测算法实现输入视频实时人脸检测和定位;通过人眼粗定位,采用hough变换实现人眼瞳孔精确定位;根据人眼瞳孔坐标实现人脸图像的几何变换、尺度归一化和直方图均衡化等预处理,实现标准人脸图像采集。其次,分析了真实人脸和欺骗人脸图像的成像模型及防欺骗检测的常用方法,确定了基于单帧人脸图像、不添加辅助设备的考勤系统防欺骗检测方案;针对单一 LBP、Fourier算法识别率不高,将LBP特征和Fourier特征进行融合作为特征提取算法;用NUAA数据库进行算法测试,测试结果表明特征融合算法具有较好的效果。然后,针对主成分分析法(PCA)、二维主成分分析(2DPCA)、人脸识别算法计算量过大,未将样本标签纳入训练,采用2DPCA-LDA特征提取算法,作为人脸特征提取算法;最后用ORL人脸数据库进行算法测试,结果表明该方法较前两种算法在识别率和识别时间上具有一定优势。最后,以PC为平台对考勤系统进行设计与开发。分析了考勤系统的设计需求,结合算法完成功能模块的开发;最后对系统进行测试,结果表明,基本达到设计要求。
[Abstract]:Compared with other biometrics, face recognition has incomparable advantages and has a wider range of applications. It is a hot research topic in machine vision. In this paper, the research and design of attendance system based on face recognition technology will be carried out, and an improved method is put forward in view of the fact that the existing face recognition system is not deceptive and vulnerable to photo and video spoofing attacks. Firstly, AdaBoost face detection algorithm is used to realize real-time face detection and localization in input video, and hough transform is used to realize the accurate location of human eye pupil by rough location of human eyes. According to the coordinate of human eye pupil, the geometric transformation, scale normalization and histogram equalization are preprocessed to realize the standard face image acquisition. Secondly, the imaging model of real face and deceptive face image and the common methods of anti-deception detection are analyzed, and the anti-deception detection scheme of attendance system based on single frame face image without adding auxiliary equipment is determined. In view of the low recognition rate of single LBP,Fourier algorithm, the fusion of LBP features and Fourier features is taken as feature extraction algorithm, and the algorithm is tested with NUAA database. The test results show that the feature fusion algorithm has good results. Then, for the principal component analysis (PCA),) two-dimensional principal component analysis (2DPCA), the face recognition algorithm is too much computation, does not include sample labels into the training, 2DPCA-LDA feature extraction algorithm, as a face feature extraction algorithm; Finally, ORL face database is used to test the algorithm. The results show that this method has some advantages over the former two algorithms in recognition rate and recognition time. Finally, the attendance system is designed and developed on the platform of PC. The design requirement of attendance system is analyzed, and the function module is developed with algorithm. Finally, the system is tested, and the result shows that it basically meets the design requirements.
【学位授予单位】：西南石油大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】