局部二元Haar特征Kadane多阈值AdaBoost面部分类识别

摘要

Aiming at the problem of feature representation and low accuracy of classification and recognition of face classification,a new algorithm called kadane multi threshold AdaBoost classifier with haar feature local binary pattern for face recognition is proposed.Firstly,it use the local binary pattern model to improve the traditional Haar features,which could improve the texture shape feature of the image model,and improve the robustness of the model to the illumination and so on;Secondly,for the single threshold weak learning algorithm can not make full use of local two yuan Haar feature information,resulting in lower classification accuracy,here proposed the multi threshold Kadane based on AdaBoost optimization,which achieve the high accuracy of the face recognition;Finally,through the experimental comparison,the proposed algorithm has a better recognition rate of more than 90％,which is better than the selection of the contrast algorithm.%针对面部分类检测识别过程中,存在的纹理形状特征表征及分类识别算法精度不高的问题,提出一种基于局部二元Haar特征表示的Kadane优化多阈值AdaBoost面部分类识别算法.首先,利用图像局部二元模式对传统的Haar特征表达形式进行改进,提高图像模型的纹理形状特征表达能力;其次,针对单阈值弱学习算法不能充分利用局部二元Haar特征信息,造成分类精度较低的问题,提出基于Kadane优化的多阈值AdaBoost分类器,实现局部二元Haar特征表示下的面部高精度识别;最后,通过实验对比显示,所提算法的面部有效识别率可达90％以上,要优于选取的对比算法.