Research on Vehicle Trajectory Prediction and Warning Based on Mixed Neural Networks

When driving on roads, the most important issue for driving is safety. There are various vehicles, including cars, motorcycles, bicycles, and pedestrians, that increase the complexity of road conditions and the burden on drivers. In order to improve driving safety, a deep learning framework is applied to predict and announce the trajectory of a car. This research is divided into three parts. Lane line detection is adopted first. Secondly, car object detection is employed. Lastly, car trajectory prediction is a key part of our research. In addition, real images and videos in the driving recorder are used to simulate the real situation the driver sees from the driver’s seat. Car detection is utilized to obtain the coordinates of the car in these images, reaching an accuracy of 0.91 and then predicting the future trajectory of the car, obtaining a loss of 0.00024 and costing 12 milliseconds. It can precisely mark the position of the car, accurately detect the lane line, and predict the future car’s trajectory. Through the prediction and announcement of the car trajectory, we verified that our model can correctly predict the car trajectory and truly enhance the safety of driving.