Real-Time Driver Gaze Direction Detection Using the 3D Triangle Model and Neural Networks
Wen-Chang Cheng, You-Song Xu · 2013
In this paper, we propose a real-time driver gaze direction detection system using a 3D triangle model and neural networks to monitor the driver's distraction. This method uses two cameras and open CV (Open Source Computer Vision Library) to locate the face, eyes and mouth position. The location of the eyes and mouth are automatically taken by two cameras. Then the information of their locations is transmitted to the triangle model, and the difference of stereo vision is calculated for depth information. Finally, we use neural networks as a classifier to accomplish our driver awareness system. The test data include 9 kinds of gaze directions, each containing 10 pictures. In the experiment result, we can achieve about an 83% detection rate base on our testing data with 9 directions. Through analyzing input face images, the model can run in real-time speed at about 30 frames per second on a normal computer.