【作者】 周琳；

【导师】 刘春；

【作者基本信息】 河南大学 ， 工程硕士（专业学位）， 2018， 硕士

【摘要】 在计算机视觉领域中,人脸检测是一个至关重要的研究方向。它利用现代先进的信息技术对人脸进行智能的检测,分析人脸包含的所有特征信息,并对这些体现人脸特点的信息进行处理。人脸检测技术如今已经应用到人们生活的各个领域内,促进了人们生活的信息化、安全化。人脸检测的研究在近几年内已取得了明显的进步。早期的研究只能够在没有任何背景的情况下检测到人脸的位置。现如今,随着深度学习等新技术在人脸检测中的应用,一些人脸检测技术已经可以非常准确地检测出在任何场景下的人脸,可以检测出多个角度拍摄的人脸图像,并可以根据检测到的人脸信息,判断两张图像中的人脸相似度,分析出人的年龄、性别和表情等。深度学习是当前人脸检测研究中广泛使用的技术。为了实现基于深度学习的人脸检测算法并验证该算法相对于传统人脸检测算法的有效性,本文的主要工作包含两部分内容:一是研究了传统的基于AdaBoost的人脸检测算法的基本原理,并基于OpenCV实现了该算法,采用AFW数据集和FDDB数据集对该算法进行试验验证分析;二是研究了基于卷积神经网络的深度学习基本原理,基于Caffe框架实现了基于卷积神经网络的人脸检测算法,最终采用了AFW数据集和FDDB数据集对该算法进行了实验验证。验证结果表明,基于深度学习的人脸检测算法不仅能够检测到正脸,还能够识别传统的基于AdaBoost算法所不能检测的侧脸,基于深度学习的人脸检测算法的准确率明显高于传统算法。更多还原

【Abstract】 Face detection is a very important research direction in the field of computer vision.It uses modern advanced information technology to detect the human face intelligently,analyzes all the feature information contained in the face,and deals with the information that embody the features of the face.Face detection technology has been applied to various areas of people’s lives,and has promoted the informatization and security of people’s lives.The research of face detection has made significant progress in recent years.Early research can only detect the location of human faces without any background.Nowadays,with the application of deep learning and other new technologies in face detection,some face detection techniques have been able to detect face accurately in any scene,detect face in the images taken from multiple angles.They can also judge the similarity of face in the two images and analyze the person’s age,sex,and expression according to the detected face information.Deep learning is a widely used technology in current face detection research.In order to realize the face detection algorithm based on deep learning and verify the effectiveness of the algorithm compared to the traditional face detection algorithm,the main work of this paper consists of two parts.First,we have analyzed the basic principles of the traditional face detection algorithm based on AdaBoost,and implemented the algorithm based on OpenCV,and verified its effectiveness based on the AFW data set.Second,we have analyzed the basic principle of deep learning based on convolution neural network,implemented the convolution neural network based face detection algorithm by using the Caffe framework,and finally verified its effectiveness based on the AFW dataset.we have also compared the performance between the AdaBoost based face detection algorithm and the deep learning based face detection algorithm.The results shows that the deep learning face detection algorithm can not only identify the positive face,but also recognize the traditional side face which can’t be detected by the AdaBoost based algorithm.,The accuracy of the depth learning face detection algorithm is obviously higher than that of the traditional algorithm.更多还原