Adaboost를 이용한 얼굴인식

Abstract

Nowadays computer just like a blindman, it only can accept the information from keyboard or mouse rather than achieving and handling message by itself from the real world. For the sake of letting computer recognize the world and getting message by itself, machine vision appears. Meanwhile, in order to achieving better performance in programs solving by computer itself, artificial intelligence develops. Now the performance of computer plays a great role in daily life, people get more and more interest in computer communication. So far, new type of machine vision doesn't depend on traditional input device. The cost performance of computers raising and cost of getting video reducing make the system of computer vision can be used in embedded system. It means that the system of computer vision can be installed in any electronic product. In the future, electronic products with high vision system will make our life more convenience. Face vision processing is an important part of computer vision processing. Face analysis includes face recognition and face detection. The purpose of face analysis is to achieving user’s identity and property (like face emotion analysis). Face recognition can be used in criminal identity recognition, bank and CIQ monitoring and so on. Hence, this paper is focused on face detection. Face detection uses a method to detect face image. Now face detection is different from previous one that detected face in simple background. For using face detection in practice, face detection system must have the ability to detect face fast and exactly. In this paper, a fast and efficient face detection method is presented which relies on the Adaboost algorithm and the features of Haar. We can calculate each part of face image's characteristic value and compare the characteristic value with non-face image's characteristic value at the same position by using rectangle features. At the same time, we can train a classifier that made up of several rectangle features which could separate face and non-face images exactly by utilizing the boosting algorithm, the process of face detection is depicted in Fig.1.