基于肤色模型和特征提取的人脸检测与分析.docVIP

PAGE PAGE 34 基于肤色模型和特征提取的人脸检测与分析 摘要 人脸作为典型的生物特征，在人脸识别、身份鉴定、智能监控、多媒体信息检索、人机交互、虚拟现实、远程视频会议等领域具有广阔的应用前景，实现这些应用的重要部分是人脸检测技术，近年来，计算机软硬件的发展，提高了人脸检测系统的性能，在学术界有理论研究的百家争鸣；在工业界也可以看到各种实用的人脸检测系统。 肤色是人脸相对稳定的重要特征，又由于人脸特征（如人眼、鼻子、嘴唇等）集中了人脸的大部分信息，在人脸图像研究领域具有重要意义。因此，本文将结合肤色模型和特征提取的方法实现人脸检测和分析的过程，具体内容包括以下2大部分。 1.本文在YCbCr色彩空间进行肤色聚类，使用高斯模型表示肤色分布，利用阈值分割实现人脸与非人脸的分类，再利用基于欧拉数的肤色区域筛选和基于模板匹配的人脸精确定位，最终实现在多人静止彩色图像的前提约束下的人脸检测及定位。 2.在人脸定位的基础上，利用面部器官的几何分布关系确定人脸特征的候选区域，在候选区域内利用局部灰度特征提取人脸特征，包括人眼，鼻子，嘴唇。 本文通过MATLAB2013b编程实现了本文的全部算法，并通过Wider Face测试图像库和自制的单人正面人脸图像库进行算法测试，从实验结果可以看出：本文算法针对背景简单的彩色正面人脸图像有较好的检测性能，同时也实现了单人正面人脸的特征提取，能够适用于特定环境下的人脸检测和分析。 关键词：人脸检测；肤色模型；特征提取 ABSTRACT Human face is a typical feature of biological, and has broad application prospects in areas of face recognition, personal identification, intelligent monitoring, multimedia information retrieval, human computer interaction, virtual reality, remote video conference and so on. And face detection is an important part to realize all the above. Recent years, with the development of computer hardware and software, the performance of face detection system has improved. There are lots of theoretical researches in academia and also many practical face detection system. The skin color is a relatively stable feature of face. And facial features, such as eyes, nose, mouth, and so on, concentrate most information of human face, which has significant meaning in the research area of face detection. Thence, this paper implemented an algorithm of face detection and analysis combining the two aspects above. The specific works are listed as follows: (1) The paper uses the YCbCr color space which has good characteristics in color clustering, and adopt Gaussian model to do the skin segmentation. Then the paper filters the candidate skin area by a test of Euler’s numbers and locates human face accurately by a algorithm based on template-matching. And the first part aims to achieve face detection and locating of colorful still image. (2)On the basis of