智能视频之人脸识别技术的分析和应用

【作者】 李勇；

【导师】 王江；

【作者基本信息】 云南大学 ， 软件工程， 2010， 硕士

【摘要】 根据公安部“城市报警与监控系统建设”工程在试点城市的推广的要求,2007年昆明市在全市开展城市监控系统建设,包括社会资源监控系统和社会治安监控系统建设两个方面。目前,随着系统建设的完成和投入使用,在公安案件侦破过程中发挥了重要作用,为刑侦工作提供了强有力的证据基础,如何更有效和更好的开发应用成为建设者后期工程着重考虑的方面。在震慑违法犯罪行为的同时,监控系统的应用可以从下面几个方面进一步深入考虑和研究:整合地理信息系统平台可以快速跟踪案件的态势和可疑人的路线;通过视频图像的智能化运用可以更为有效的防范案件的发生,对自然人的行为进行识别可建立预警机制;对于自然人其面部特征建立信息采集并进行比对筛选可疑人等等都是今后系统功能的完善和发展方向。本文通过对智能视频监控技术的研究和分析,对比了常见的几种生物识别模式,选取了人脸识别作为研究对象,介绍了常用的人脸识别的实现方法,并且分析了其原理,重点是在现有视频监控基础上实现对人脸检测和人脸识别技术进行模型构建,实现Adaboost快速人脸检测方法,分析了该方法所使用的矩形特征及积分图的计算方法,利用这些矩形特征构造了若干个弱分类器并级联成一个强分类器,设计了人脸检测系统,图像经过检测的同时,对检测到的人脸进灰度变化、直方图均衡等预处理,使得图片更适合用于人脸识别;采用基于主成分分析的特征脸分析算法,完成人脸训练和人脸识别两个主要步骤,把人脸检测和人脸识别两个部分有机的结合起来,构建一个自动完成人脸检测和识别的较为完整的系统原型。最后文章从公安工作的应用角度出发,分析了当前公安工作中智能分析技术的重要性以及对人脸检测和识别技术的需求,对人脸识别技术系统的应用做了分析和思考。更多还原

【Abstract】 According to Ministry of Public Security "Urban alarm and monitoring system construction" project in the pilot cities to promote the requirements of the city of Kunming in 2007 to carry out construction of urban monitoring systems, including the social resources, monitoring systems and social security are two aspects of monitoring system construction. Now, with the completion of system construction and put into use, in the police solved the case and make an important role in the process for the criminal investigation has provided a strong evidence base, how to be more effective and better development and application to become builders of the post-project focus on take into account aspects.Deter illegal and criminal acts in the same time, monitoring the application of the system from the following aspects can be further in-depth consideration and study: Integrating geographic information system platform can fast track the situation and cases of suspicious routes; through the intelligent use of video images can be more for the effective prevention of cases occur, the behavior of natural persons may be established to identify early-warning mechanism; for a natural person whose facial features of the establishment of information collected and compare the screening suspicious person, and so are all the future functions of the system improvement and development.In this paper, for intelligent video surveillance technology research and analysis, compared several common biometric model, selected face recognition as an object of study, describes the implementation of commonly used face recognition methods, and analysis of its principles, focusing on In the existing video surveillance based on the human face detection and face recognition technologies for modeling, to achieve rapid Adaboost face detection methods, analysis of the methods used by rectangular features and integral image is calculated using these rectangular structure features a number of weak classifiers and cascading into a strong classifier, the design of human face detection systems, image tested at the same time,on the detected face into the gray-scale changes, such as histogram equalization pre-processing, makes the image more suitable for face recognition; principal component analysis based on the characteristics of the face analysis algorithms, to complete the training and face recognition face two major steps to face detection and face recognition combining the two parts of the organic build an auto-complete face detection and recognition of a more complete system prototype.The article from the angle of job applications, analyzed the current of public work of the intelligence analysis and technical importance to detect and identify the face of recognition technology system application for analysis and consideration.更多还原