基于改进Gentle-Adaboost算法的疲劳驾驶中人脸检测研究  

Analysis of Face Recognition Based on Improved Gentle-Adaboost Algorithm

作　　者：蔡伽 马镜璇 武卫东[1] 王建霞[1] CAI Jia;MA Jing-xuan;WU Wei-dong;WANG Jian-xia(School of Information Science and Engineering, Hebei University of Science and Technology, Shijiazhuang 050018, China;School of Cyber Security and Computer, Hebei University, Baoding 071000, China)

机构地区：[1]河北科技大学信息科学与工程学院,河北石家庄050018 [2]河北大学网络空间安全与计算机学院,河北保定071000

出　　处：《数学的实践与认识》2018年第10期89-96,共8页Mathematics in Practice and Theory

摘　　要：疲劳驾驶检测中人脸检测的算法有很多，其中相对成熟的是Gentle-Adaboost此算法虽然识别率较高，但是识别时间较长，研究在识别过程中使用均值哈希算法进行算法优化，并引入缓存的概念来缩短识别时间．首先，根据Gentle-Adaboost算法和Haar特征来提取基本特征点，然后通过对比前后两帧图像的哈希指纹来减少人脸检测次数，最后通过缓存数据库来存储相似图片的哈希指纹，运行一段时间后仅需通过对比哈希指纹就能精准的找到人脸区域．通过实验，可以证明改进后的平均识别时间可减少原时间的80％．There are many algorithms for face recognition in fatigue driving detection. Among them, Gentle-Adaboost is relatively mature. Although the recognition rate is higher, the recognition time is longer. In this paper, we use the mean hash algorithm to optimize the algorithm and introduce the concept of caching to shorten the recognition time. First, according to Gentle-Adaboost algorithm and Haar feature to extract the basic feature points, and then the number of face recognition is reduced by comparing the hash fingerprints of the two images before and after finally the hash fingerprint of similar images are stored by the cache database, after running for some time, just by comparing the hash can be accurately found in the face area. Experiments show that the improved average recognition time can be reduced by 80% of the original time.

关 键 词：Gentle—Adaboost算法 HAAR特征 疲劳驾驶 均值哈希算法 人脸检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]