基于肤色与改进的Adaboost的人脸检测

Face Detection Based on Skin and Modified Adaboost Algorithm

针对传统的Adaboost算法检测速度块,准确率低的问题,提出一种基于肤色与改进的Adaboost算法以提高人脸检测的正确率。该算法首先利用肤色特征快速排除绝大部分背景区域,然后在肤色区域中利用快速积分图提取人脸的Haar特征,接着使用阈值设定的方法对传统的Adaboost算法进行改进,并将每次检测的最优分类器级联形成最终的强分类器,通过强分类器对Haar特征判别,检测出人脸部分。实验证明该算法有效的提高了人脸检测的准确率。

In order to solve the low accuracy ot the detection speed in the traditional AaaDoost algonthm,thiS paper presents a modified Adaboost algorithm based on skin color to improve the accuracy of the face detection.First,the algorithm rules the out the most background area quickly by using the skin features.And the Haar feature of the face is pick up by the using the rapid integration figure in the skin color of area.Then the threshold setting method improves the traditional Adaboost al gorithm.The optimal classifiers from every detection are cascaded the ultimate strong classifier.According to the Haar char acters distributions from the strong classifier,the face area is detected.