基于分块LBP融合特征和SVM的人脸识别算法

Face recognition algorithm based on block LBP fusion feature and SVM

针对传统局部二值模式(LBP)特征提取方法在光线和人脸表情变化的情况下表现不佳、单一方法提取出的特征不能多角度表现出整张人脸的特征信息的问题,提出一种基于分块LBP融合特征和支持向量机(SVM)的人脸识别方法。先将人脸图像划分为不同的块,对每一块提取LBP特征;然后将不同分块的像素均值特征与LBP特征进行融合,用融合后的特征向量来表征人脸;最后引入SVM作为分类器对上述特征进行分类。在YALE、ORL标准人脸库以及自建人脸库上进行实验验证,实验结果表明:该方法识别准确率分别能达到95. 15%,99. 75%,96. 25%,对比实验显示,该方法优于C4. 5决策树、随机森林等传统方法。

Aiming at the problem that the traditional local binary pattern(LBP)feature extraction method does not perform well under the condition of changing of light and facial expression,and the feature extracted by a single method cannot express the feature information of the entire face from multiple angles,an approach of face recognition based on block local binary pattern(LBP)fusion feature and the support vector machine(SVM)is proposed.For each face image,it is divided into several blocks,and the LBP features are extracted from each block,then pixel means in different blocks are fused with LBP features.The fused feature vector by all the blocks including LBP features and average pixel values are used to represent the whole face.Finally,the SVM is introduced and used as classifier to classify the above features.Experiments are carried out on YALE,ORL and self-built face database.It turns out that the recognition accuracy can respectively reach 95.15%,99.75%and96.25%.Comparative experiments show that this method is superior to the traditional methods such as C4.5 decision tree and random forest.