The MERL system uses 3D face models and specifically targets illumination and pose variation, which they think as the most critical factors limiting performance.

MERL系统使用三维人脸模型，重点解决他所制识别率两个因素，光照和姿势变化。