Face Detection Research Based On The Adaboost Algorithm And Different Color Space

Face detection has been a more and more popular method for face recognition today. Face recognition has been used in many places such as image retrieval, video monitoring, military base and other sensitive areas of access control. There are three key steps in face recognition:face detection,feature extraction and feature matching.Face detection has been the first and the most important step in the whole process of face recognition.So, how to make face detection quickly and efficiently will directly affect the whole process of face recognition.According to the research background mentioned above, I have done some further research on the first step of the face recognition in this paper.and I combine some popular methods which are considered as good ones and make some improvements in order to make the improved one can do a better job than before.By improving the methods,we can recognize the pictures that we want much more precisely,and at the same time,we can do some further research on the application fields of the face recognition.The method I talk about in this paper is a combination of two methods, one is the method based on the method of Adaboost,the other is the method based on skin color detection,and I do some improvement on the latter.At first,I introduce the two methods above,and in the process of the skin color detection, three different skin color models are set up in RGB?HSV and YCbCr color space, at the same time,I also make comparisons among them.Finally, I choose the YCbCr color space as a skin color modeling space in this paper. But there are also some defects in the process of YCbCr color space,for not all the pictures that I put in can get ideal detection results.So based on this,I do some improvement in this paper,that is,I change the parameters of the transformation matrix which is used in the transformation from RGB color space to the YCbCr space.By this,the improved method can make a better choice for us and recognize the ideal picture for us precisely.In this paper,I improve the method introduced above and make a subsequent completement in order to make the skin area which I choose is much more similar to human face skin color area,and I use a great amount of pictures to test the system I established. Test results show that the system I established can recognize the ideal picture for us precisely.What's more,this improvement can promote the the popularity of face detection and will be better for the information security and maintenance of us.