摘要作为自动人脸处理系统的第一步, 人脸检测是机器视觉与模式识别领域的研究热点,具有重要的学术价值。人脸识别或辨认、人脸定位以及人脸追踪等都与人脸检测密切相关。局部二值模式(local binary pattern,LBP)是一种灰度范围内的纹理描述方式, 本论文中采用了一种基于改进的LBP算子的侧面人脸检测算法,首先将检测的图像进行预处理。为了检测到不同大小的人脸,采用窗口滑动技术。将提取到的LBP图像直方图特征用于SVM样本训练。再采用滑动窗口技术扫描图像窗口得到LBP特征,再用SVM进行分类判断。实验结果初步证实,该方法可以有效的检测出侧面人脸。64525
毕业论文关键词 LBP 均匀模式 窗口滑动技术 SVM 直方图 特征判断
毕业设计说明书(论文)外文摘要
Title LBP-based Face Detection
Abstract As the first step of automatic face processing system, face detection is the research focus of machine vision and pattern recognition and has important academic value. Recognition or identification, face orientation and face tracking and face detection are all closely related. Local Binary Pattern (local binary pattern, LBP) is a grayscale texture description method within the scope of this paper uses a method based on improved LBP operator side face detection algorithm, first detected image pre-treatment. In order to detect faces of different sizes, using the sliding window technique. Will be extracted to the LBP histogram features for SVM training sample. Then scanned using a sliding window technique to get LBP feature image window, and then determine the SVM classification.
Preliminary experimental results proved that this method can effectively detect the side of the face.
Keywords LBP Uniform mode Sliding window technique SVM histogram feature to determine
目 次
1 引言 6
1.1人脸检测 6
1.2研究现状与方法概述 8
1.3待解决的问题 9
2 LBP算法 10
2.1 LBP算子及其改进 10
2.2 改进的LBP算子 13
2.3 LBP图像的特征提取(feature extraction) 15
3 SVM特征分类 17
3.1 SVM 17
3.2 径向基核函数 (Radial Basis Function)–RBF 19
3.3 用OpenCV中SVM算法的大致流程 20
4 实验 21
4.1侧面人脸检测方法性能评估 21
4.2 检测器 22
4.3 实验结果 22
4.4 实验结果分析 27
结 论 29
致 谢 30
参考文献 31
附录A 核心LBP算法的实现 33
图 1 人脸分析流程 7
图2 人脸的遮挡、不同表情、图像的质量、旋转灯等都会影响人脸检测 8
图3:基本的LBP算子操作过程 11
图4 基本LBP算子对图像处理前后的对比