Research on Cross-Age Face Verification Based on Artificial Neural Network Under Examination Environment

Nowadays, the fairness of the examination has attracted wide attention, so it's meaningful to design a system to ensure the fairness of examination.Most of current methods focus on detecting abnormal behavior in exams, but our idea is to detect surrogate exam-taker(ghost writer) by comparing the live face photo taken by the camera outside of the examination room with the photo of identification card——one can enter the examination room only if both are identical.In this paper, we have two contributions.Firstly, we design an end-to-end learning framework by designing a multi-task deep neural network for cross-age face verification under examination environment, the proposed CNN network can be directly applied in feature extraction of the face photos and ID-Cards photos.Secondly, we adopt our designed model into the data set from The Authority of Anhui Education, the data set contains 380,000 photos from 190,000 persons, our model achieves superior performance in comparison to current methods, we reach the accuracy at 99.3%.