基于深度学习的汽车防追尾预警系统设计OA北大核心CSTPCD

In allusion to the current mainstream rear end collision prevention systems,there is a common problem of single identification factors and insufficient prevention effect on the driver themselves.A deep learning based automotive rear end collision prevention warning system is designed,which can integrate YOLOv5 pruning technology,attention mechanism,PID optimizer and other methods into the training of network models to optimize model accuracy and reduce model volume.The rear end risk of vehicles is calculated mainly based on distance judgment,and supplemented by speed,acceleration,and accident hazard judgment.After each warning,the data is uploaded to the Internet of Things platform by means of the MQTT protocol,and safety analysis is conducted on the driver when the system ends running.The system is ultimately deployed in the TensorRT environment for the further optimization.The experimental results show that the designed car rear end collision prevention warning system has fast response speed,strong adaptability,and accurate risk assessment.