人脸识别毕业设计

信息科学与技术学院毕业论文课题名称：基于特征识别的人脸检测系统学院：信息科学与技术学院完成日期：二○一七年五月十九日摘要I摘要我的毕业设计题目是基于特征的人脸检测系统，这个系统不仅仅能够检测人脸，还具有识别人脸的功能。检测人脸检测部分的算法采用的是于仕祺老师的LBP特征加GentleAdaBoost分类器相结合的算法，提取识别特征部分的算法采用的是Google在2015年提出的基于深度学习策略的一种人工神经网络FaceNet，较为新颖，其准确率高，在光照不足，姿态和表情变化剧烈时仍能保持稳定，具有很强的鲁棒性。该系统的界面使用MFC编写，在具体实现中了应用了多线程编程技术实现了一个简单的生产者消费者模型，从而提高了系统的识别效率，另外，对人脸的识别模块还使用了Python，C++混合编程技术引入了Google的开源深度学习框架Tensorflow作为对FaceNet的具体实现，数据库使用的是SQLServer2012，连接数据库使用的是微软公司的ADO组件。该系统主要有信息采集模块和实时监控模块两个部分，前者完成对任务样本的信息采集工作，后者完成在实时监控的情况下对出现在画面中的人脸进行检测和识别，检测部分的速度可以达到40~60的FPS，识别部分由于计算量较大，只能达到2~5的FPS。该系统经过简单的硬件支持和部署之后，基本可以完成在实际场景中的简单应用，具有一定的学术研究和实际应用价值。关键词：人脸检测；人脸识别；机器学习；Tensorflow；实时监控ABSTRACTIIABSTRACTThetopicofthisgraduationprojectisFaceDetectionSystembasedoncharacteristicswhichachievesthefacedetectionandfacerecognitiontwofunctions.ThealgorithmoffacedetectionpartusesakindofenhancedalgorithmbasedonLBPfeatureandGentleAdaBoostclassifierproposedbyShiQiYu,thealgorithmofextractingfacefeatureusedinrecognitionpartusesakindofmanualneuralnetworkFaceNetbasedondeeplearningstrategyproposedbyGooglein2015.FaceNethasreachedhigharruracyanditisrobustnesstothechangeofillumination,postureandexpression.TheinterfaceofthissystemifwritteninMFC,andinrealimplementation,theapplicationofmulti-threadedprogrammingtechnologyrealizesasimpleproducerandconsumermodelwhichacceleratethewholerecognitionefficiencyofthesystem,inaddition,therecognitionpartalsousesthePython,C++mixedprogrammingtechnologywhichintroducesGoogle’sopen-sourcedeeplearningframeworkTensorflowasaconcreteimplementationofFaceNet,thedatabaseisusingSQLServer2012,thelinkofdatabaseusesMicrosoft’sADOcomponents.Thesystemconsistsoftwoparts:theinformationcollectionmoduleandthereal-timemonitoringmodule,theformercompletestheinformationcollectionofhumansamples,andthelattercompletesthedetectionandrecognitionofthefacesthatappearinthepictureinthecaseofreal-timemonitoring.Thespeedofdetectionpartcanreachesto40~60FPS,therecognitionpartcanonlyreachto2~5FPS,forthelargecalculationcost.Aftersomesimplehardwaresupportanddeployment,thesystemcanbeusedinactualsceneforsimpleapplicationwhichhascertainresearchandpracticalapplicationvalue.KEYWORDS:Facedetection;Facerecognition;Machinelearning;Tensorflow;Real-timemonitoring石河子大学信息科学与技术学院毕业论文1目录1绪论.................................................................31.1课题............................................................................................................................................31.2课题背景.....................................................................................................................................31.3课题研究目的及意义.................................................................................................................31.3.1研究目的........................................................................................................................31.3.2研究意义........................................................................................................................41.4国内外研究现状.........................................................................................................................51.4.1国外................................................................................................................................51.4.2国内................................................................................................................................61.5设计时间.....................................................................................................................................71.6内容及分工.................................................................................................................................71.6.1内容................................................................................................................................71.6.2成果................................................................................................................................72理论和技术...........................................................82.1理论............................................................................................................................................82.1.1检测部分的LBP特征+GentleAdaBoost分类器.......................................................82.1.2识别部分的GoogleFaceNet.....................................................................................112.2技术..........................................................................................................................................152.2.1MFC简介......................................................................................................................152.2.2Tensorflow简介........................................................................................................172.2.3ADO组件简介..............................................................................................................183需求分析及概要设计..................................................193.1需求分析......................................................................