Research And Implementation Of Age Estimation Based On Facial Feature

Automatic age estimation based on facial images is an important but challengingtopic in human face recognition. It has both theoretical and practical significance,especially in image and video retrieval, advertising distribution and other fields.To establish a face database that contains the age information is the first step of thestudy. So far, the international public databases such as FG-NET and MORPH are allwesterners’ face databases, and some of the images are in poor quality and thedistribution of age of them are very uneven. Therefore, it’s not suitable for Asians ageestimation.In this paper, we built a big Asians face database. We mainly studied on humanfacial image feature extraction and age estimation model, and proposed an ageestimation method based on facial features. Firstly, we need to preprocess on a faceregion for wipe out the influence of lighting and position. Then we need to do thefeature extraction including the geometric proportion features and the texture featureswhich base on the ASM and LBP respectively. Thirdly, in order to reduce the dimensionof the texture and improve the computing speed, we need to do the principle componentanalysis. Finally, we use SVM to do the classification.Our method can estimate the age intervals for an input testing facial image withsmall errors. Experimental results show that the methods combining the geometricproportion features and the texture features can obtain better experimental results.