基于肤色及改进的Adaboost算法的人脸检测

为提高人脸检测的实时性和准确率,提出一种基于肤色与改进的Adaboost算法结合的人脸检测新方法。首先在YCbCr色彩空间下建立混合高斯肤色模型处理待检测图像,分割出肤色区域得到候选人脸区域。然后通过Harr矩形特征扩展与样本权值更新改进Adaboost算法,进行人脸检测。实验表明,该方法较好地处理了复杂背景下彩色图像人脸检测的漏检、错检问题,提高了检测速度和精度。

Research on Face Detection Based on Skin Color and Improved Adaboost Algorithm

In order to improve the accuracy and speed of face detection,the article brings about a new way of face detection which combines skin color and improved Adaboost arithmetic.First,a Gaussian skin model is set up to handle the pending images under the YCbCr Color Space so as to obtain the Color likelihood images,and the color area is divided by adaptive threshold method to obtain the candidate face area.Then the article improves the Adaboost arithmetic by Harr rectangle feature expanding and sample weights update.This method solves the problems of fallout ratio and false dismissal probability of face detection in complicated background images,thus improving the detection efficiency.