Study On Visualization Of Vehicle Assisted Driving Based On Augmented Reality

At present,vehicle-assisted driving technology is an important research direction for improving traffic safety.The vehicle-assisted driving visualization technology is a research hotspot in this direction.Augmented reality technology can intuitively display vehicle-assisted driving information to drivers to avoid distraction and reduce traffic accidents.Therefore,the research on vehicle-assisted driving visualization based on augmented reality has important theoretical significance and engineering application value.The main research work of the dissertation includes the following items:1.The overall framework of the vehicle-assisted driving information visualization system was designed to capture the real 3D scenes through the camera of the smart mobile device.The 3D registration and tracking technology was used to perform 3D environment reconstruction and camera pose acquisition.Based on these information,the virtual navigation information and vehicle information was accurately added to the real scene,and the final processed visual image was displayed on the screen of the mobile device and presented to the users.2.A 3D augmented reality registration and tracking method based on the combination of orbital fast and rotated brief(ORB)feature points and simultaneous localization and mapping(SLAM)was studied.When the tracker considers the current frame to be a fixed map,the backend optimization uses a local beam adjustment method to optimize the key frames within the local window.This method allows the local beam adjustment method to be completed in a fixed time.Real-time calculations are finally achieved by detecting large maps and optimizing lightweight pose corrections to detect large closed loops.3.A new IMU initialization method was studied.In the inertial vision SLAM algorithm,tracking and local beam adjustment optimization methods both require a fixed state,therefore accurate IMU initialization is needed to provide accurate state estimation.This paper solved the problem through the divide-and-conquer method.First,we used monocular vision ORBSLAM algorithm to estimate the poses of some key frames and unknown scale factors.Then we computed the gyro bias through the key frames with the same direction,and finally we solved the scale and the direction of gravity.Finally,we solved scale and gravity direction,using the known gravity amplitude value to solve the acceleration deviation,adjusting the scale and gravity direction,and extracting the speed of all key frames,which only requires sensor motion so that all unknowns can be observed without any conditional assumptions.4.An augmented reality assisted driving visualization system based on a mobile platform was constructed,and related software was developed.The software has the function of superimposing navigation and vehicle information into a real 3D environment which can intuitively provide drivers with useful driving information.At the same time,the experiments verify that the visual inertial 3D registration tracking algorithm proposed has better drift elimination ability and higher accuracy.