摘要
针对人脸深度图像的分类识别问题展开研究,提出一种自适应3DLBP(3D Local Binary Pattern,3DLBP)特征提取算法。该特征提取算法以机器学习理论为基础,首次将反馈学习理论与3DLBP特征提取过程相结合,以保证特征提取算法对训练样本集的变化具有理想的普适性;同时,为了提高自适应特征提取算法的稳定性,提出使用多分类器对反馈学习过程进行优化。实验结果表明,自适应3DLBP特征对训练样本集的变化具有较好的有效性和稳定性,在FRGCv2.0人脸数据库上取得了理想的识别效果。
The face recognition based on intensity imagesis,mainly introduces an adaptive 3DLBP（3D Local Binary Pattern）feature extraction algorithmis is proposed.On the basis of machine learning theory,this adaptive feature extraction algorithm for the first time combines feedback learning with 3DLBP feature extraction,and this combination can guarantee the algorithm has ideal common adaptability for the changes of sample sets.At the same time,complementary multiple classifiers are used in pre-classification process in order to improve the robustness of the algorithm.Experiment results confirm that the obtained adaptive 3DLBP features has ideal common adaptability for changes of sample sets,and achieved good test results on FRGCv2.0 database.
出处
《电视技术》
北大核心
2013年第19期46-49,共4页
Video Engineering
基金
湖北省科技攻关计划项目(2006AA301B44)
