摘要
基于传统LBP特征人脸检测方法的不足,提出针对LBP特征描述方法进行改进,建立LBP特征金字塔,调整LBP的特征描述方式使其在多尺度的图像中有较好的描述能力。并且设计Real Adaboost分类器实现对改进LBP特征的分类,为人脸检测提供了一种新的方法。实验结果表明,基于改进后的LBP特征人脸检测方法的最大检测率为94.1%,高于Haar算法的92.8%和传统LBP算法的93.2%,所以改进后的算法对人脸具有更好的描述和鉴别能力。
Based on the deficiency of traditional LBP feature face detection method,this paper proposes to improve the LBP feature description method,establish the LBP feature pyramid,and adjust the feature description of LBP to make it have better ability to describe it in multi-scale image. And the Real Adaboost classifier is designed to improve the classification of LBP features,which provides a new method for face detection. The experimental results show that the maximum detection rate based on the improved LBP feature face detection method is 94. 1%,which is higher than 92. 8% of the Haar algorithm and 93. 2% of the traditional LBP algorithm. Therefore,the improved algorithm has a better description of the human face and the ability to identify.
出处
《信息技术》
2018年第2期1-4,10,共5页
Information Technology
基金
国家自然科学基金(61170200)
江苏省重点研发计划(社会发展)项目(BE2015707)
