Transfer Learning-Based Vehicle Collision Prediction

Traffic accident is an important problem in modern society. Vehicle collision prediction one of the key technical points that must be broken through future driving system. However, due to complexity traffic environment and difference emergency ability drivers, it very difficult predict vehicle collision. Although experts scholars have tried monitor accidents real time according environmental conditions, overly agile warning or inaccurate may cause serious consequences. Therefore, order more accurately occurrence collision, this paper analyses models mode based on transfer learning using previous performance data vehicle, so as situation even vehicle. Finally, a real-world Internet Vehicles set, implements large number experiments verify effectiveness proposed model.