Research On Vehicle Detection And Trajectory Prediction Based On Surveillance Video

With the continuous development of China's economy and the progress of science and technology,the monitoring of the traffic network is becoming more and more perfect.The traffic video monitoring equipment has been deployed at many intersections,which provides data support for the research of vehicle detection,trajectory prediction and many other technologies.Through these technologies,the characteristic information in the traffic network can be mined,which provides suggestions for improving the traffic and formulating traffic strategies It can be used as a supplement.Vehicle detection is a common technical means in traffic video processing,and the technical difficulty is how to solve the influence of wrong detection,missing detection and external environment changes on the detection results.To solve these problems,this thesis proposes an improved ViBe algorithm.According to the application of vehicle trajectory prediction in traffic,this thesis proposes a vehicle trajectory prediction method based on improved Hidden Markov Model.Vehicle lane changing is a common form of mobility in the process of vehicle driving,and it is easy to scratch or collide in the process of lane changing.In order to ensure the safety of lane changing process,this thesis studies the vehicle collision warning technology,and designs a reasonable warning mechanism by using the collision time.The main work of this thesis is as follows:The dynamic target vehicle is detected.Aiming at the poor performance of traditional target detection algorithms in dynamic target detection,In this thesis,an improved ViBe algorithm is proposed to realize the dynamic target vehicle detection.The experimental results show that the proposed algorithm has higher detection accuracy in dynamic environment;The vehicle trajectory is predicted.Traditional trajectory prediction methods based on kinematics model and machine learning model only use the current state information of vehicle,which is easy to cause large errors.To solve this problem,this thesis proposes a vehicle trajectory prediction method based on Improved Hidden Markov Model.The track sequence information is used to predict the vehicle trajectory,and the appropriate learning method is used to train the model.The experimental results show that the model has higher prediction accuracy and can be used for long-term prediction;Design vehicle collision warning mechanism.Aiming at the problem that the design of early warning mechanism based on collision time TTC will ignore other factors,this thesis proposes the risk factor R combined with collision time TTC and vehicle braking time TTA to quantitatively analyze the severity of collision risk and design a reasonable collision early warning mechanism.The experimental results show that the warning mechanism designed in this thesis can improve the driver's driving safety.