基于Genlte boosting算法的人脸检测与识别研究-计算机软件与理论专业论文.docxVIP

摘 要 人脸自动识别技术指用计算机分析人脸图像，从中提取有效信息，实现身份自动 鉴别的技术。与其他生物特征识别相比，它更直接、更友好、更自然，使用者无任何心 理障碍。 本文在前人研究工作的基础上，对图像预处理、特征提取、分类器训练等方法进 行了深入研究，针对人脸的特点，设计并改进了一些有效的算法来解决人脸检测和识 别中遇到的问题。 (1) 在图像预处理方面，本文在前人提出的Gentle boosting学习算法的基础上，增 加了在算法训练阶段以及特征融合阶段，对输入图像首先进行直方图均衡处理，有效 改善图片质量，突出图像中的有用信息，更有利于特征提取: (2) 在特征提取方面，采用基于小片的区域特征提取的方法，以小片的形式在图像 上随机采集，计算其与图像中心点的位移。与其他方法相比，本方法不拘泥于人脸的 轮廓形状及灰度突变的硬性分布规定，复杂环境下的图像也能全面地采集到特征: (3) 采用主元分析法进行人脸识别，提出首先对图像进行直方图均衡化处理及边 缘检测，对人脸检测的结果进行识别，通过实验验证这两种方法结合起来的可行性: (4) 实现一个人脸检测识别系统，使用一个自制的人脸库，在该系统上对上述算法 进行了测试。实验结果验证本文所提算法运行效果良好，达到了设计的目的。 关键词:人脸检测; 人脸识别; 特征提取; 机器学习. 万方数据 I – Abstract The automatic face recognition is a technology that using computer to analyze the human face images and extracting effective features from the human face images, then to recognize them. Compared with other biometric methods, face recognition is more direct, friendly, nature, and users can use it without any inconvenience. Based on the research of forerunners, along with the research in the algorithm of face detection and recognition, we have a deep research in image preprocessing, features extraction, and training the classify. Aiming at the characteristics of human face, several effective algorithms have been designed in this thesis to solve the problems in face detection and recognition. First of all, in the image preprocessing, this study basing on Gentle boosting learn- ing algorithm put forward by people before, adds a method for histogram equalization of the images input in the algorithm training stage and feature computing stage, which can effectively improve the quality of the images, strengthen the useful information of the images and be more benefit to extract feature farther. Secondly, in the feature extraction, the study uses a method based on patch part- based feature extracting, which gather the features of the images random in type of patch and compute the distance to the center of the image. Compared with others, this method can also gather the features round