- 上一篇：MSP430的串口转以太网的系统的设计
- 下一篇：基于PRI的脉冲信号分选算法的仿真与应用
摘 要人脸识别技术的飞速发展使人们生活中的很多事情都变的智能化了。基于高性能嵌入式的实时人脸识别系统，以嵌入式开发板作为系统设计基础，实现了实时人脸识别的功能。人脸识别技术的首要工作是人脸检测，需要设计出鲁棒性好、效率高的人脸检测方法。人脸检测的目的是确定在包含人脸的图像区域中所检测到的人脸个数等等。
本文研究的内容是以CUDA平台为开发环境，GPU并行化计算为方法，设计出基于高性能嵌入式的实时人脸识别系统。为了能够实现系统的功能，本系统主要采用了Haar分类器的方法对人脸图像进行一系列处理，对图像处理算法优化，加快图像的处理速度，有效地实现了系统的实时性。
本文以人脸识别系统的研究与发展现状为开端，随后对系统进行了整体的需求分析。完成系统的整体设计，并对图像处理模块进行详细分析。调用OpenCV人脸识别库中的人脸识别算法，完成对人脸的检测与识别。最终，完成了对系统的运行并将系统的性能进行了各方面的详细分析。lf0076
关键字：人脸识别技术；嵌入式；CUDA平台并行化；Haar分类器；OpenCV
Abstract
The rapid development of human face recognition technology has made many things intelligent in people's life.The real-time face recognition system is based on high performance embedded developing board .Face recognition reflects the real time in the operation of the whole system. To the face recognition technology the first thing is to detect face, which requires a good and efficient face detection method. The aim of face detection is to determine all the image regions that contain the human face etc.
The content of this paper is based on the CUDA platform for the development environment and the method is GPU parallelization calculation, and then to design a real-time face recognition system based on high-performance embedded. To realize the function, this system mainly adopts the Haar classifier method to process a series of the face images and optimizes the image processing algorithm. It accelerates the speed of the image processing and then realizes the real-time performance of the system effectively.
In this paper, the research and development of the face recognition system is the beginning, and then it is the whole analysis of the system needs. It completes the overall design of the system and the image processing module for a detailed analysis. It calls the face recognition algorithm in the OpenCV face recognition database to complete the detection and recognition of the human face. Finally, it operates the system and analyzes the performance of the system in detail.
Keywords: Face detection and recognition; Embedded; CUDA platform parallelization; Haar classifier; OpenCV
目 录
第一章 绪论1
1.1 引言1
1.2. 本文的目的及意义2
1.3. 国内和国外研究情况及发展的趋势3
1.4. 本文的主要工作4
第二章 系统分析5
2.1 研究目标5
2.2 本课题整体功能需求分析5
2.3 系统设计模块需求分析6
2.4 本章小结7
第三章 人脸识别方法介绍8
3.1 概述8
3.2 CPU和GPU异构平台的并行化8
3.3 人脸图像预处理9
3.3.1 均值滤波算法和高斯滤波算法9
3.3.2 图像的灰度化处理10
3.3.3 图像增强和图像复原11
3.4 人脸检测12
3.4.1 基于模板的人脸检测方法12
3.4.2 基于特征的人脸检测方法12
3.4.3 Haar分类器13
3.4.4 AdaBoost算法13
3.5 本章小结14
第四章 人脸识别系统的实现15