A Study And Application Of Age Estimation Based On Face Image

With the development of computer technology, image processing technology and human-computer interaction have become one of the hot research in computer science. Using face image as an example, the face image contains a wealth of information, people's gender, age, facial expression, health information can be obtained from the face image. The study of the age estimation through facial image get more and more attention in recent years. Predicting the person's age accurately and beginning a more in-depth study, can change people's habits and the way of life. It has great significance in the information time.This thesis mainly studies the feature extraction and fusion based on the FG-NET facial age library. In the feature extraction stage, in order to overcome the influence of light, this thesis use uniform Local Binary Pattern(uniform LBP) and Canny edge detection operator. In order to adapt to the study of face age characteristics, this paper try to use the threshold adjustment in the Canny edge detection. In the feature fusion stage, this thesis uses two strategies——the direct fusion strategy and the weighted fusion strategy. This thesis studied the age feature fusion through the two strategies.In the classification stage, this paper uses the principal component analysis method(PCA) for image data reduction work. Then, using the improved support vector machine(SVM) to learn and estimate the age.The experiments show that this method has good effect on age feature extraction. Compared with similar studies, this fusion strategy shows better in age feature and predicts higher accuracy. In the practical application stage, this paper designed a system to push the information service for people through the age estimation. The experimental results show that the proposed method can overcome the influence of the illumination on the age estimation, and obtain a more stable prediction effect.