Research Of Face Detection Based On Skin Color And AdaBoost Algorithm

With the development of Computer Science,face recognition gets more and more attention.As the first step of face recognition,face detection has become one of the most active topics for computer vision and pattern recognition..In this thesis,dissertate the basic concepts of face detection technology,analysis and discuss basic theory on a variety of methods of face detection,focusing on the color image of the face detection,adopt face detection method based on color segmentation and AdaBoost technology.The experiments indicate that the methods of face detection proposed in this paper are reasonable,showing a certain degree of theoretical and practical value.The research work of this paper mainly includes the following several aspects:First,face detection algorithm based on the model of faces' skin color is carried out by using the skin color information of color image.A simple Gaussin probability model based on YCbCr color space is used,with this model,skin color is calculated of similarity and binary segmented,Then size and length-to-width geometrical analysis about object region is carried out.Finally,Remove non-human face region by image-filled and regional certification, detected face effectively.Secondly,using face detection method based on AdaBoost learning algorithm,which fast calculate haar-like features by "integral figure" method,to build a robust cascade classifier.Focusing on the disadvantages of classical AdaBoost algorithm,this paper analyses the issues of degeneration and distortion of sample weights in training process and come up with a new method to avoid the phenomenon of degeneration in a certain extent.The experimental results show that the new method effect is remarkable.Finally,as skin color detection achieves high false acceptance rate in color image with complex background and AdaBoost algorithm does not work well in multi-pose and multi-face images,a novel face detection method combined skin color detection and AdaBoost algorithm is proposed in this paper.The experimental results show that the new scheme is able to detect faces with high detection rate and low false acceptance rate and performance better than skin color detection and AdaBoost algorithm.It can be effectively applied to the cases of mufti-pose and mufti-face images.