Research On Real-time Rear-end Collision Warning Mechanism Of Internet Of Vehicles

Recent years,with the continuous development of internet of vehicles,autonomous driving technology has been widely used to improve the safety of vehicles in smart cities,and is being carried on more and more vehicles.In the real road environment,rear-end collision is one of the main causes of traffic accidents.The combination of an efficient rear-end collision warning strategy and an emergency braking system can greatly improve the safety of road traffic.Building a reliable real-time back-end collision warning strategy is particularly important for the popularization of autonomous driving technology.Many researchers have done a lot of work on collision prediction which can be divided into two categories: traditional parameterized methods and machine learning-based methods.In recent years,machine learning-based methods have proved to be more suitable for solving this complicated prediction problem than parameterized methods in practice.However,due to limitations in feature extraction and collection of real dataset,machine learning methods face great challenges in back-end collision prediction.To solve these problems,this thesis proposes a novel back-end collision prediction strategy(RCPM)based on convolutional neural network,which applies the convolutional neural network to the field of vehicle collision detection.In this thesis,gram matrix is used to convert the vehicle trajectory data based on time series into a three-dimensional image as the input of the neural network.In addition,a novel synthetic oversampling method based on genetic algorithm is used to expand the number of minority samples to solve the category imbalance of the dataset.In this thesis,a neural network model is constructed in conjunction with dilated convolution,and the receptive field is increased exponentially without increasing the resolution of the feature map.This work shows that,RCPM effectively improves the accuracy of collision prediction on the basis of ensuring real-time prediction.Through the rear-end collision warning algorithm proposed in this article,the safety of autonomous driving can be effectively guaranteed,which is that autonomous driving has a stronger adaptability under real road environment.