Research On Face Detection And Face Alignment

Automatic Face Recognition (AFR) aims at endowing computers with theability to identify different human beings according to face images. Such a researchhas both significant theoretic values and wide potential applications. After more than30 years' development, AFR has made great progress especially in the past ten years.The state-of-the-art AFR system can perform identification successfully underwell-controlled environment. However, evaluation results and practical experiencehave shown that AFR technologies are currently far from mature. A great number ofchallenges are to be solved before one can implement a robust practical AFRapplication. This article mainly discusses the problem of face detection based on skincolor and the problem of facial point location based on statistical learning methods.As the earlier period of the face recognition, face detection is very important.This thesis first describes some main current methods of face detection, and thenintroduces the knowledge of color dimension. A face detection method based onskin-color model and template matching is discussed, this method mainly use theskin color model to get the skin region and then carry on the confirmation using thetemplate match. This method overcome the disadvantage of template matchingmethod and improves the precision and speed of the detection.Face alignment should be carried out after face detection, and then this thesisdiscusses the face alignment algorithm using Active Shape Model as emphasis. PointDistribution Model is described firstly, and three aspects of ASM are discussed:aligning the training set, modeling shape variation and local grey model. Then thisthesis describes the whole process of active shape model and some improvements ofthe active shape model.At last, a simulation system of face detection and alignment is realized, and lotsof test is carried out. The result of the experiment shows that this system is effective.