基于LDP特征和贝叶斯模型的人脸识别

发布时间：2018-05-05 04:18

本文选题：人脸识别 + 局部方向模式　；参考：《计算机科学》2017年12期

【摘要】：针对现有的局部方向模式LDP(Local Directional Pattern)方法仅利用了图像自身的LDP特征的问题,提出将LDP特征直方图与贝叶斯(Bayes)模型相结合的方法,从而有效使用人脸图像的先验信息,以提高人脸的识别率。第一步,在相互独立的训练集上,学习同类样本图像和异类样本图像的LDP直方图特征相似度的先验信息,并估计类条件概率密度函数(同类样本与异类样本分别进行计算);第二步,利用人脸图像的LDP直方图来比较该图像是否为某一类型图像的概率数值大小;第三步,使用贝叶斯规则进行分类。仿真结果证明,在ORL库与Yale库上,与传统PCA,LBP和LDP算法相比,所提方法得到的人脸识别率均有显著提升。
[Abstract]:In order to solve the problem that the existing local direction pattern LDP(Local Directional pattern method only utilizes the LDP features of the image itself, a method combining the LDP feature histogram with the Bayesian Bayes model is proposed to effectively use the prior information of the face image. In order to improve the face recognition rate. The first step is to learn the prior information of LDP histogram feature similarity between the same kind of sample image and the heterogeneous sample image on the independent training set. And estimate the class conditional probability density function (the similar sample and the heterogeneous sample are calculated separately; the second step, using the LDP histogram of the face image to compare the probability value of the image whether the image is a certain type of image; the third step, Use Bayesian rules to classify. The simulation results show that compared with the traditional PCA-LBP and LDP algorithms, the proposed method can significantly improve the face recognition rate in the ORL library and the Yale library.
【作者单位】：兰州理工大学计算机与通信学院;
【基金】：国家自然科学基金:多民族欠发达地区传染病传播动力学特征分析与建模(61263019)资助
【分类号】：TP391.41

【相似文献】