基于分块加权LBP技术的人脸识别算法

The Face Recognition Algorithm Based on Block Weighted LBP Technology

人脸识别属于生理特征识别,是基于人的脸部特征信息进行身份识别的生物识别技术,是我国人工智能技术领域的首个成熟技术。LBP(Local Binary Patterns)算法,又称局部二值模式算法,是一种灰度范围内的纹理描述方式。传统LBP算子提取的特征信息只能体现局部的人脸信息,不能完整表达全部人脸信息。在基本LBP算法的基础上提出基于分块加权LBP技术的人脸识别算法,将人脸分为5×3子分块,根据人脸五官在人脸识别中的不同贡献度赋予不同的权重提取人脸信息特征。通过在ORL和YALE两种人脸数据库中训练不同样本数,比较传统LBP方法、5×3分块LBP方法和5×3分块加权LBP方法的人脸识别准确率,实验证明分块加权LBP技术在人脸识别中可以有效提高识别准确率。

Face recognition technology belongs to the recognition of physiological characteristics, and is a biometric identification technology based on human face characteristic information. It is the first mature technology in the field of artifi- cial intelligence technology in China. LBP algorithm, also called local binary pattern algorithm, is a texture description method in gray range. The feature information extracted by the traditional LBP operator can only reflect the local face informa- tion, and can not express the whole face information. Based on the basic LBP algorithm, a face recognition algorithm based on block weighted LBP is proposed, which divides the face into 5 x 3 sub-blocks. According to the different contribution of facial features to face recognition, different weights are given to extract face information features. Different samples are trained in ORL and YALE face databases. The accuracy rate of face recognition based on the traditional LBP method is compared with that of the block LBP method and the weighted LBP method. Experiments show that block weighted LBP technology can effectively improve the accuracy of face recognition.