选择性眩光防避系统保护下的司机面部检测

来车发出的强光及低位置太阳光引起的眩目效果会严重危害驾驶安全性，眩光防避系统可智能地在司机的眼睛区域投下阴影，以保护眼睛免受眩目效果的影响.然而，在眼睛被保护性阴影覆盖时，系统仍需对司机脸部进行检测，以实现进一步的监控.针对该问题，采用部分覆膜训练（PMT）、连续亚块训练（CST）以及交叠亚块训练（OST），提出了阴影遮挡下的司机人脸检测算法；同时提出了将司机脸部轮廓检测及头部旋转角度估计方法运用于眩光防避系统，以增强复杂光照条件下的司机脸部检测精度与鲁棒性.实验结果验证了所提算法的有效性.

Driver face detection under protection of the selective dazzling avoidance system

Dazzling effects caused by strong light from oncoming vehicles and low-position sunlight were seriously harmful to driving safety. The dazzling avoidance system could intelligently cast a shadow on the driver''s eye region to protect the eyes from dazzling effects. However, when the eyes were covered by protective shadows, the system still needed to detect the driver''s face to maintain further monitoring. The driver''s face detection algorithm under occlusion were proposed by employing partially masked training (PMT), consecutive sub-block training (CST) and overlapped sub-block training (OST). Meanwhile, the methods of profile face detection and rotation angle estimation the of driver''s head were proposed to enhance the accuracy and robustness of the driver face detection under the condition of complex light. The experimental results verified the effectiveness of the proposed algorithm.