Research On The Head Up Display System Based On Three Dimensional Line Of Sight Tracking

With the popularity of the automobile and the rapid development of artificial intelligence,the interactions between human and the car are more and more frequent in the driving process.At present,the interactions between the human and the car are mainly based on the touch and voice.However,due to the limitations of dialect,voice recognition cannot be universal.At the same time,voice control has a certain delay,making it difficult to apply to the automotive system.In the process of driving,the driver needs to control the steering wheel in real time,however,in touch interactive mode,the human hands will leave the steering wheel,result in a certain driving safety risks.Eye control technology,with no limitations of the dialect,frees hands to a certain extent and avoids the security problems caused by touch interaction during driving.This thesis introduces the eye control technology into the car head up display(HUD),which can provide a better interactive ability for HUD.The eye control method uses 3D gaze tracking algorithm and dynamic blink recognition to obtain human eye movements,simulates the fall point of sight into the touch screen click position,simulates the two consecutive conscious blink into the click into of touch screen,simulates the gaze into the selected target and moves combine with the fall points of sight,blink again represent the release of the target.The use of transparent display,the results will be displayed on the front of the car to ensure that the human eye does not leave the front,and then design a vehicle HUD interaction system of human and vehicle.There are two key technical problems to be solved in the eye controlled HUD system,eye tracking and blink recognition.In line of sight tracking module,this thesis presents a method of 3D vision estimation based on Intel Realsense 3D camera,which can solve the problem of line of sight estimation due to head movement.In the line of sight estimation,first of all,uses two lines of sight to mark the spots,and the use of Intel Realsense SDK to provide the face of the mark points,tracks the spot by calculating the distance between the mark points of the face which provided by Intel Realsense SDK and the spot.Then,finds the world coordinates of the pupil through the pupil recognition algorithm.Finally,Compares and analyses the results of the line of sight estimation algorithm with the actual measurement results.In the blink recognition module,this thesis puts forward the use of the two eyes of the eye and the distance between the upper and lower eyelids to determine whether the change in the blink of an eye.This dynamic recognition method can identify the blink of the eye by real-time tracking of the eye state,and has a high stability.In the process of recognition,the image processing algorithm is used to collect data,such as grayscale,two value,and projection and so on.And designs an improved two value algorithm,which can be used to extract the image features according to the characteristics of the pixels.Finally,tests and analyses the denoising algorithm and recognition method.The results show that the algorithm can adapt to the change of illumination to a certain extent,and preserve the eye information needed by the blink detection.