Research On Prediction Of Vehicle Trajectory And Early Warning Of Violation Of Right Turn Behavior At Intersections

With the vigorous development of Internet of Things technology and mobile Internet technology and the widespread use of traffic video surveillance,massive trajectory data has been generated.The prediction of vehicle trajectory can be achieved by mining the laws contained in the trajectory data,which has a wide range of application prospects.In recent years,with the rise of artificial intelligence,vehicle violation detection systems have been gradually put into application,but the current detection systems only have detection functions,that is,identify and punish vehicles after violations.If the vehicle trajectory is predicted and then the violation judgment is made,the early warning of vehicle violations can be realized,which is beneficial to control the violations from the source.First,the behavior of vehicles at intersections was investigated,and the results showed that the incidence of right-turning with compacting lane and right-turning on the straight lane was the highest.Therefore,these two violations were selected as the research objects,and then the centralized investigation sites were selected to collect videos and perform the original video.After stabilization and segmentation,the trajectory data is extracted,and the vehicle behavior at the intersection is analyzed based on the trajectory data,thereby screening the trajectory data for prediction.Secondly,select the LSTM model in the deep learning method to establish a trajectory prediction model.The whole process includes the preparation of the basic environment and the construction of a multivariate LSTM model.The multivariate LSTM model is constructed based on Python,and the corresponding evaluation index value is output after repeated training..Thirdly,the particle swarm algorithm is used to improve the basic LSTM model,and the trajectory prediction model is established based on the improved LSTM model.At the same time,the traditional machine learning model is used as the comparison model.Finally,the error index values of the three models are compared,and the results show the improvement The LSTM model is optimal.Finally,early warning of illegal right-turning behaviors,including two modules,trajectory prediction and violation judging,is based on the range of vehicle trajectory coordinates.If the coordinate value is within the set range,it can be judged as an illegal right-turning vehicle.After confirming that the accuracy of the judgment result is higher,the early warning is completed;finally,the intelligent early warning idea is proposed,and the vehicle that is judged to be in violation of the rules will be warned by the cloud system.If the violation is still violated after the warning,it will be punished accordingly.The research in this paper has certain theoretical significance for the trajectory prediction problem and the development of intelligent transportation systems.It provides new ideas for the vehicle trajectory prediction problem at intersections,and the proposed early warning of vehicle illegal right-turning behavior can promote illegal right turn to a greater extent.The reduction of behavior improves the efficiency of road traffic management and the safety and order of roads.