The Research On Driver's Head Pose Estimation

With the economy and society developing, driving car has entered people's daily life and became one of the most important choices for outside. At the same time, traffic safety gradually arouse the people's attention. Vehicular safety relies on the ability of people to maintain constant awareness of the environment as they drive. As new vehicles and obstacles move into the vicinity of the car, a driver must be cognizant of the change and be ready to respond as necessary.A lot of research work has been done in the field of head pose estimation. There are several kinds of ways for this research. The hybrid method, which is one way among them, can combine the advantages of different methods in order to improve the accuracy.To give necessary remind for drivers in driving process and avoid the driver causing traffic accidents because of drunk driving and fatigue driving. This paper provides a method combine the geometry method with the nonlinear regression, which is used for driver head pose estimation successfully. The algorithm extracts the feature points of eyes, mouth and other parts of the head, then uses these feature points to build geometric models. Finally, using nonlinear regression analyzes the geometric model data and determine the motion of the head. The whole process is divided into three main parts by this method:1) Firstly, getting the video information and the corresponding image information, taking the normalization for the images and using the SIFT algorithm to extract the feature points of the images of the driver's face;2) Secondly, making clustering analysis for the image feature points by fuzzy c-means algorithm, calculating the distance between the corresponding feature points;3) Thirdly, using nonlinear regression based on neural network to train the distance between the clustering centers, mapping the data to the linear space can be divided and estimating the driver's head pose.Experimental results show that the proposed method can improve the validity and accuracy of driver's head posture estimation under the condition of guaranteeing the prediction accuracy, and has high robustness. It is an effective and safe method to apply it effectively in the course of driving assistance systems to judge the driver's head pose.