Research And Application Of Head Pose Tracking Technology Based On Computer Vision

Head pose estimation based on computer vision has attracted more and more attention as an important research topic in the field of natural human-computer interaction.So far,head pose estimation has been successfully applied to many important fields,such as automobile safe driving,entrance guard detection,video conference,identity identification and so on.This paper deeply studies the methods of head pose estimation.Firstly,this paper proposes the method of combining skin color model with gray integral projection to detect facial feature points.Secondly,the face alignment algorithm based on SDM model is optimized,and the head pose is calculated according to the change of facial feature points.Finally,this paper implements the application of head pose estimation in automobile safe driving.The main contents and innovations of this paper are as follows:1)In this paper,the head pose is determined by the position change of facial feature points in video images,so the accuracy and efficiency of face feature detection is very important.After studying the mainstream methods of face feature point detection,this paper proposes a method to combine skin color model with gray integral projection.This method detects the feature points region by skin color model,and then uses the gray integral projection method to locate the feature points accurately.The method reduces the detection range of feature points,so the detection efficiency is improved,at the same time,this method normalizes the image and processes it in mathematical morphology,thus reducing the influence of illumination,expression changes and other factors on the experimental results.2)This paper studies the face alignment algorithm based on SDM model,and it is found that there are some problems such as the large amount of feature points,calculation and the difference of face shape in the measurement process.To solve these problems,this paper optimizes the algorithm.This paper adds the initialization function after facial feature points detection and the initial face and feature points are used as training samples.After the optimization of the algorithm,the sample image and the image to be tested are the same person,which eliminates the influence of the facial shape difference on the experimental results.It also reduces the dimension of feature points,reduces the computational complexity and improves the computational efficiency.At the same time,the sample feature points need not be manually labeled,thus reducing the training time and improving the experimental efficiency.3)This paper implements the application of head pose estimation in automobile safe driving.This paper carries out fatigue detection and concentration test by analyzing the change regulation of the driver's head posture,and designs a vehicle safety driving assistant system based on computer vision.The experiment shows that the system can accurately detect the fatigue state and the mental state of the driver in real time so it has certain practical value.