毕业设计：基于树结构的人脸检测算法的研究及实现.docx

PAGE 15 PAGE 基于树结构的人脸检测算法的研究及实现 摘 要 人脸检测算法最早起源于人脸识别问题，也是人脸识别系统的关键步骤。人脸检测是一种能够在图像中确定是否存在人脸，并返回人脸在图像中的位置和大小等相应信息的技术手段。人脸检测算法种类较多，但大都存在一定的局限性，特别是当人脸受外界因素干扰或自身发生姿态旋转变化时，会使得人脸检测的难度加大，检测效率特别是准确率有所降低。而基于树结构的人脸检测算法融合多种人脸检测算法方案，结合树结构的特殊性实现对人脸检测算法的优化，并建立统一的人脸检测、姿态估计和关键点标记模型，进而实现人脸检测在效率和准确率上的提升。 本论文主要针对基于树结构的人脸检测算法技术展开研究，在Matlab中编程实现这一人脸检测系统。通过对具有代表性的Viola- Jones人脸检测算法的研究，实现两组人脸检测算法的实验对比。参考评价标准确定基于树结构的人脸检测算法在人脸检测效率和准确率上存在优势，并将这一算法应用在大规模人脸图像数据中，为该算法在生活场景中的应用提供论据支持。 关键词：人脸检测，树结构，LFW人脸数据库 The Research and Application of Face Detection Algorithm Based on Tree Structure Abstract The issue of face detection, derived from face recognition problem, is a significant procedure of face recognition system. Face detection is a kind of technical method that is capable of determining whether human face exists in the image, and returns the information of faces such as positions and sizes. Face detection has many kinds of methods: while these methods all have their own limitations, especially when faces have interferences of external environment or have posture and rotate diversities, the difficulty of detection would increase much and the accuracy would correspondingly decrease. However, Face detection algorithm based on tree structure merges several kinds of face detection algorithms, and combined the particularity of tree structure to achieve the optimization. Simultaneously, the algorithm established a unified model of face detection, pose estimation and landmark localization, thus improved the efficiency and accuracy of face detection. This thesis mainly researches on face detection algorithm based on tree structure, and the application of the face detection system in Matlab. With researching the Viola- Jones face detection algorithm, we confronted two face detection algorithms in order to determine the former face detection algorithm makes great improvements in efficiency and accuracy of face detection. Hence, by utilizing the algorithm in large-scale face image databases, we can p