Skin Age And Age Classification Based On Facial Features

As the age grows,the face is deformed due to the movement and growth of the bones;the skin is gradually losing its vitality,mostly with fine lines and spots.Skin age analysis and evaluation based on facial feature in the field of computer vision,has more and more attention.The automatic skin age assessment and the scientific quantification of skin health status have great research significance and application value in many industries such as beauty,medical care and cosmetics.Nowadays,the research on skin age at home and abroad is mainly concentrated in the field of clinical medicine and Cosmetology,most of the research needs to rely on laser scanning confocal microscopy or skin care and other professional equipment.Based on the digital camera face image to skin aging research is very rare.The age information becomes an increasingly important new biological feature.Age classification based on facial features has gradually become one of the hot research topics in the field of computer vision.In this project,we analyze the face images of digital cameras and mobile phones,and analyze the facial features,focusing on the growth model of skin age with aging,and the age classification based on face images.The main research contents and results are as follows:1.Face image preprocessing technology is studied,including pixel gray scale,image denoising,face alignment and face semantic segmentation,for image feature extraction,content understanding,skin age and age classification research;2.The model of skin age change based on visual cognition is proposed.Through the statistical analysis of the bottom of the pixel characteristics,high-level semantic features of facial skin,The model of facial skin age based on visual cognition is established.Effectively quantify and verify the facial skin with age,smoothness is getting lower and more wrinkles;3.The 2-off age classification model more suitable for human visual cognition is proposed.Based on the SVM classifier and the low-level feature vector,the 2-off age classification model is proposed.Compared with the model before improvement,the accuracy of age classification is improved obviously.4.A loss function based on age series is proposed.By improving the loss function of SoftMax layer,making full use of the sequence information between ages,to achieve a 13% accuracy rate increase.