Construction Of Time-space Four-dimensional Joint Imaging Model And Location Prediction Of Moving Objects

With the rapid development of the information society,the motion target positioning prediction technology has created great application value in the field of intelligent driving.The traditional moving target positioning prediction method is mainly based on the kinematic model of the three-dimensional world,which is highly dependent on the environment.Few studies regard the time parameter and the space parameter as a whole.In order to solve the above problems,this paper proposes to treat "time-space" as a whole,construct the model of time-space four-dimensional joint imaging system,and study the calibration method of time-space four-dimensional system.The method based on artificial neural network provides a solution to the nonlinear mapping problem of complex system.The main research content of this paper is as follows:(1)Establishment of time-space four-dimensional joint imaging system model.In this paper,based on the binocular sequence image and the central perspective imaging model,the spatial object relationship is expanded along the time axis into a time-space four-dimensional imaging system,and a visual imaging system model with time dimension is established.(2)Research on calibration method of time-space four-dimensional system.Based on the traditional camera calibration method,time and motion parameters were added to expand and derive the mathematical model,and the calibration method of time-space four-dimensional joint imaging system was proposed,which laid a theoretical foundation for the subsequent experiments.(3)Establishment of time-space four-dimensional positioning prediction model based on BP neural network.The fitting prediction relation of time-space four-dimensional model is established by BP neural network.Through the self-built sample database data set,the neural network is trained and learned,and the trained and stable model is used to verify the physical motion positioning prediction experiment.(4)Application experiment of pedestrian motion positioning prediction in driving scene.The prediction model proposed in this paper is applied to indoor environment and outdoor environment respectively,and the positioning prediction of pedestrians in front of the vehicle is carried out in three kinds of motion models: fast walking,slow walking and running.The self-built sample database is used to train the prediction model and predict the track position information of pedestrians at some point in the future.The prediction results are used to determine whether pedestrians can cross within a safe time range.In this paper,the prediction model is verified by the guide rail in indoor environment.Then,the short distance pedestrian movement position prediction experiment is carried out in the laboratory environment.Finally,the pedestrian movement positioning prediction experiment is carried out in the outdoor environment.Experimental results show that the time-space four-dimensional positioning prediction model proposed in this paper can learn the four-dimensional time-space object-image mapping relationship well.Although there are still errors in the prediction model,in the early stage of this study,experiments can prove that the model is effective and has engineering significance,which provides complete,reliable and accurate data support for judging the motion state of the dynamic target ahead at a certain time in the future during the intelligent driving process.