# 一种改善人脸识别效率的快速筛选方法研究Study on the Faster Novel Screening Technology for Improving the Efficiency of Face Recognition

Abstract: Face recognition is an important identification technology and has a wide application prospect. Nowadays, how to make the face recognition faster and more accurate is one of the pursuing targets in this field. The target can be achieved through a local significant feature and effectively reducing the number of comparisons. Currently, most of the face recognition methods are used to extract the features of the entire face image, and through one-by-one comparisons with the images in the database to obtain final results. In this study, a screening technique is proposed that can effectively improve the defects of over-compared features and excessive comparing times. In order to design this screening technology, the influence of locally significant features on the recognition rate is explored by using variance analysis to obtain the optimal screening technology process. The studied results show that to compare with the current methods, the proposed technology not only can maintain the same recognition rate, but also can improve the recognition efficiencies upon 115.5% and 52.9% on the recognition time subject to the face databases of Extended Yale Face Database B and MECL, respectively.

1. 引言

2002年汤姆克鲁斯主演的电影《关键报告》中，街头识别系统随时扫描识别过往路人身份的情节，正逐渐在你我的生活中上演。由于人脸识别具备远距离运作非接触式的特性，为人带来的便利更胜于其它生物识别技术，举凡阿汤哥电影《不可能的任务》中的虹膜识别，或是警方办案常使用的指纹识别等，因此，当其技术藩篱被突破时，随之而来的应用不计其数。根据知名市场研究公司MarketsandMarkets推估，人脸识别市场产值在五年内可望以13.9%的年均复合增长率(Compound Average Growth Rate, CAGR)成长，由2017年的40.5亿美元跃升至2022年的77.6亿美元。

2. 人脸识别系统

Figure 1. Face recognition process

3. 人脸数据库介绍

3.1. Extended Yale Face Database B

Extended Yale Face Database B是由38个人在64个不同照明条件下拍摄9种姿势而成，并且通过光源方向与中心相机轴之间的角度(12˚，25˚，50˚，77˚，90˚)将数据库分为5个子集，共2414张图像。本文采用数据库中每个人的第一张人脸图像作为等待比对的人脸数据库，共38张图像，并将子集1~3的所有正面图像作为测试图像，共1174张图像。

3.2. 自制人脸数据库

4. MATLAB识别系统建立

4.1. 实验设备

4.2. 预处理

Figure 2. Image preprocessing process

$I\left(x,y\right)←\frac{I\left(x,y\right)}{{\left(mean{\left(|I\left({x}^{\prime },{y}^{\prime }\right)|\right)}^{\alpha }\right)}^{1/\alpha }}$ (1)

$I\left(x,y\right)←\frac{I\left(x,y\right)}{{\left(mean{\left(\mathrm{min}\tau |I\left({x}^{\prime },{y}^{\prime }\right)|\right)}^{\alpha }\right)}^{1/\alpha }}$ (2)

$I\left(x,y\right)←\tau \mathrm{tanh}\left(\frac{I\left(x,y\right)}{\tau }\right)$ (3)

Figure 3. The effect of Image preprocessing process (left), the original image after preprocessing the image (right)

4.3. 特征提取

Figure 4. LBP operation way

Figure 5. Feature extraction

4.4. 变异数分析

Table 1. Variance with configuration check (features)

Table 2. One-way anova (features)

Table 3. Games-Howell multiple comparison

Table 4. Descriptive statistical analysis (features)

4.5. 分类识别

${\chi }^{2}\left(p,q\right)=\underset{i}{\sum }\frac{{\left({p}_{i}-{q}_{i}\right)}^{2}}{\left({p}_{i}+{q}_{i}\right)}$ (5)

Figure 6. Screening technology process

Table 5. Screening technology promote efficiency

5. 结论

[1] (2017) Facial Recognition Market by Component (Soft ware Tools and Services), Technology, Use Case (Emotion Recognition, Attendance Tracking and Monitoring, Access Control, Law Enforcement), End-User, and Region-Global Forecast to 2022.
https://www.marketsandmarkets.com/

[2] Kaya, Y., Kayci, L. and Uyar, M. (2015) Automatic Identification of Butterfly Species Based on Local Binary Patterns and Artificial Neural Network. Applied Soft Computing, 28, 132-137.
https://doi.org/10.1016/j.asoc.2014.11.046

[3] Tang, Z., Su, Y.C., Er, M.J., Qi, F., Zhang, L. and Zhou, J.Y. (2015) A Local Binary Pattern Based Texture Descriptors for Classification of Tea Leaves. Neurocomputing, 168, 1011-1023.
https://doi.org/10.1016/j.neucom.2015.05.024

[4] Zhang, Y. and Hua, C.J. (2015) Driver Fatigue Recognition Based on Facial Expression Analysis Using Local Binary Patterns. Optik, 126, 4501-4505.
https://doi.org/10.1016/j.ijleo.2015.08.185

[5] Alelaiwi, A., Abdul, W., Solaiman Dewan, M., Migdadi, M. and Muhammed, G. (2016) Steerable Pyramid Transform and Local Binary Pattern Based Robust Face Recognition for E-Health Secured Login. Computers and Electrical Engineering, 53, 435-443.
https://doi.org/10.1016/j.compeleceng.2016.01.008

[6] Huang, G.B., et al. (2008) Labeled Faces in the Wild: A Database for Studying Face Recognition in Unconstrained Environments. Workshop on Faces in Real-Life Images: Detection, Alignment, and Recognition.

[7] Ira, K.-S., et al. (2016) The MegaFace Benchmark: 1 Million Faces for Recognition at Scale. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, 27-30 June 2016.

[8] Tan, X.Y. and Triggs, B. (2010) Enhanced Local Texture Feature Sets for Face Recognition under Difficult Lighting Conditions. IEEE Transactions on Image Processing, 19, 1635-1650.
https://doi.org/10.1109/TIP.2010.2042645

[9] Xu, Y., Zhang, Z., Lu, G.M. and Yang, J. (2016) Approximately Symmetrical Face Images for Image Preprocessing in Face Recognition and Sparse Representation Based Classification. Pattern Recognition, 54, 68-82.
https://doi.org/10.1016/j.patcog.2015.12.017

Top