The Research And Implementation Of Face Detection Based On AdaBoost Algorithm

With the development of intelligent and the important of identity has become increasingly prominent since the1990s,face detection as the first step of face analysis is becoming to be a hot topic. Well, face analysis is using method of pattern recognition to get information from image of human faces, such as expressions and identity.Face detection is to distinguish whether there have faces in the video or images to be detect. If we find the target, then give the basic information of the human face back, such as the location and attitude.Firstly, focuse on the key technology of face detection. After doing a lot of research, we found that the most effective method-AdaBoost Algorithm, and it has the top evaluation rate. The algorithm is more faster and has higher detection rate. This paper choose the algorithm to be the basic theory, then found out some point which needed to improve.Secondly, as we know that, illumination problem took an important part in the system. From learning the paper [40], we found that it is far from being enough for only use HE to do pre-process. In this paper,we use gamma correction to deal with the illumination problem, and then we use plenty of experiments.Thirdly, study on the detection based Skin Chrominance Models, and found out the short of it. Do research of AdaBoost Algorithm and how to realize it. The important is to find the short comings and improve it. In my paper, firstly, I added some new haar-like features, improved the type of updating weight, and introduce cost factor into the algorithm. Using the method above, the different samples would be gave different weight, so that can improving use ratio of positive samples. What’s more we use each of the strong classifier to be the first weak classifier in the next stage, which can drop more non-human faces at the very beginning. It also will improve the speed of detection.Then, we make a comparison between the two algorithm. Afer doing plenty of experiments using the standard test library,and give several typical situations to analysis.From the result, we got the average detection and the false detection rate.The experiment result showed that the false detection rate was lower and the detection rate has also been improved. From the result we also verified the feasibility of AdaBoost algorithm using in face detection, and it can be applied in the face analysis which has complex background, also do some contribution to the investiga-tion of human faces in area such as criminal investigation and customs.