Detection of Train Driver's Pupil and Eye Corner Based on Coarse to Fine Positioning

This paper proposed a train driver's pupil and eye corner detection method based on coarse to fine positioning. First, a facial landmark positioning and tracking technology based on supervised descent method was used to roughly locate the eye corners of drivers, and extract eye images based on the position of corresponding points resulted from the correct location of facial feature points. Then, a circular sliding template was used to traverse the eye image to obtain the gray ratio of each pixel, and a weighted integral projection method was used to detect the pupils, roughly. Finally, the detected landmarks on eye corners and pupils were adopted as the initial points, and a location technology based on local binary feature was used to precisely locate the position of the pupils and eyes corners. Image and video datasets were adopted to evaluate the proposed method. The experimental results show that this algorithm can effectively locate the position of the pupils and eye corners of drivers and outperforms other state-of-the-art methods for pupil detection.