Automobile Virtual Driving System Design Based On Real Scene Modeling

With the swift growth of the amount of motor vehicles and drivers,the training quality of driver's school training institutions is increasingly demanded by the society.At present,the training institutions of driving schools adopt the way of coach demonstration and accompanying driving.This training method has many problems,such as high human cost,high time cost,on-the-spot driving impacts on traffic safety,student backlog,uneven coach teaching level and so on.Compared with the traditional driving training methods,the analog driving system has the advantages of absolute safety,energy saving,environmental protection,no environmental impact,low human cost and high learning efficiency.However,compared with actual driving,the biggest problem of simulated driving is that the screen display effect is far from the real operation effect,and the hot virtual reality technology can solve this problem to a large extent.In view of the existing problems in the current driving training industry and the shortcomings of the current driving simulator,an automobile virtual driving system based on real scene modeling is designed.The system uses real recorded panoramic video pictures as material to build the spherical driving scene of the system,and assisted by virtual reality equipment,so that the gap between the user's driving experience and the actual driving experience is greatly reduced.Firstly,the virtual driving control of the vehicle is realized by changing the pose of the vehicle model and adjusting the sphere scene picture dynamically.Secondly,the Faster R-CNN depth learning framework based on Sparse Residual Network(SRN)is adopted to detect the target of the dynamic scene picture and obtain the position coordinates of the vehicle and pedestrians in the scene picture.In view of the fact that the depth information of the object in the real world is lost in the panoramic video shot,this paper adopts the method of data regression to establish the spherical scene by analyzing the mapping relationship between the height of the vertical pixel in the picture and the distance of the object in the real world,and between the width of the horizontal pixel in the picture and the width of the object in the real world.Finally,the collision judgment function of the virtual driving system is realized by calculating the distance information of multiple targets in the scene in real time.In addition,this paper introduces the implementation of virtual driving system based on real scene modeling in Unity3 D,including sphere scene building,control system building and collision judgment,and tests and analyses the system.In the experiment,all functions are running normally,the visual and sensory effects are realistic,and the expected goals are achieved.Of course,the system is still in the experimental research stage,and there is still room for improvement and potential for improvement from the practical application.