Face detection and alignment method for driver on highroad based on improved multi-task cascaded convolutional networks

Driver's face detection and alignment techniques in Intelligent Transportation System (ITS) under unlimited environment are challenging issues, which are conductive to supervising traffic order and maintaining public safety. This paper proposes the improved Multi-task Cascaded Convolutional Networks (ITS-MTCNN) to realize accurate face region detection and feature alignment of driver's face on highway, predicting face and feature location via a coarse-to-fine pattern. Moreover, the improved regularization method and effective online hard sample mining technique are proposed in ITS-MTCNN method. Then, the training model and contrast experiment are conducted on our self-build traffic driver's face database. Finally, the effectiveness of ITS-MTCNN method is validated by comparative experiments and verified under various complex highway conditions. At the same time, average alignment errors on left eye, right eye, nose, left mouth as well as right mouth of the proposed technique are performed. Experimental results show that ITS-MTCNN model shows satisfied performance compared to other state-of-the-art techniques used in driver's face detection and alignment, keeping robust to the occlusion, varying pose and extreme illumination on highway.