嵌入式人脸检测与识别系统的研究与实现

发布时间：2018-08-07 14:50

【摘要】：人脸识别,特指分析比较人脸视觉特征信息进行身份鉴别的计算机技术。人脸识别也是模式识别与人工智能的研究热点之一。它在访问控制、司法应用、电子商务和视频监控等领域都有广泛的应用。传统的高性能自动识别系统大多基于PC架构的应用,而在实际应用中却对自动识别系统要求更高的便携性和易用性。尤其近年来,随着智能家居的兴起,嵌入式系统的快速发展,嵌入式智能视频监控系统的重要地位越来越凸显出来,它结合了嵌入式设备、人脸自动检测、自动识别系统的优点,可广泛应用于各类门禁系统、家庭安防、智能小区等场合。本文对基于OpenCV的嵌入式人脸识别系统进行了研究与实现。本文首先研究分析了一些目前流行的人脸检测和识别方法,并且深入探讨了OpenCV中Viola-Jones人脸检测法和PCA特征脸人脸识别法,通过OpenCV提供的工具函数并利用采集到的人脸样本训练了人脸正面强分类器。通过定制Linux内核和文件系统结合GT2440开发平台构建了嵌入式Linux操作系统,把OpenCV库移植到了嵌入式操作系统中,并对OpenCV库中匹配已知人脸的过程进行了优化。硬件方面通过ZC301视频采集设备与GT2440开发平台组成人脸图像采集装置,并通过以太网连接服务器方式组成人脸检测识别系统。软件方面采用多线程、模块化设计,完成了视频帧采集模块、人脸图像帧筛选模块、嵌入式系统通信模块和服务器人脸识别模块等四部分的模块设计。论文的最后对本文系统设计所完成的内容进行了总结,指出本系统的缺点和不足,提出了改进的方向。
[Abstract]:Face recognition, especially the computer technology of analyzing and comparing face visual feature information for identity identification. Face recognition is also one of the hotspots in pattern recognition and artificial intelligence. It has a wide range of applications in access control, judicial applications, electronic commerce and video surveillance. The traditional high performance automatic recognition system is mostly based on the application of PC architecture, but in the practical application, the automatic recognition system requires higher portability and ease of use. Especially in recent years, with the rise of smart home and the rapid development of embedded system, the important position of embedded intelligent video surveillance system is more and more prominent. The advantages of automatic recognition system can be widely used in various access control systems, home security, intelligent residential areas and other occasions. In this paper, the embedded face recognition system based on OpenCV is studied and implemented. This paper first studies and analyzes some popular face detection and recognition methods, and deeply discusses the Viola-Jones face detection method and PCA feature face recognition method in OpenCV. The strong face classifier is trained by the tool function provided by OpenCV and the collected face samples. The embedded Linux operating system is constructed by customizing the Linux kernel and the file system combined with the GT2440 development platform. The OpenCV library is transplanted to the embedded operating system, and the process of matching the known faces in the OpenCV library is optimized. In the aspect of hardware, the face image acquisition device is composed of ZC301 video acquisition device and GT2440 development platform, and the face detection and recognition system is made up of Ethernet connection server mode. The software is designed by multithreading and modularization. Four modules including video frame acquisition module, face image frame screening module, embedded system communication module and server face recognition module are designed. At the end of the paper, the contents of the system design are summarized, the shortcomings and shortcomings of the system are pointed out, and the direction of improvement is put forward.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.41;TP368.1

【相似文献】