现代电子技术2025，Vol.48Issue(22)：179-186,8.DOI:10.16652/j.issn.1004-373x.2025.22.029
基于改进YOLOv8的车辆碰撞预警方法研究
Research on vehicle collision warning method based on improved YOLOv8
摘要
Abstract
In allusion to the vehicle collision accident on the traffic road,a method of vehicle collision warning based on improved YOLOv8 is proposed.The improved YOLOv8 object detection algorithm is used to realize the real-time detection and tracking of the vehicle in front.The more accurate collision warning area is constructed based on the lane lines on both sides and the warning line in front,and this area is calculated and updated in real time according to the change of vehicle speed.A collision warning strategy following the principle of defensive driving is proposed by combining with the actual driving scenario,which can send out three-level warning signals of yellow,blue and red according to the relative position between the vehicle and the warning area and its movement trend.The experimental results show that the average accuracy(mAP)and frame number per second(FPS)of the improved YOLOv8 algorithm on the self-built dataset reach 89.3%and 92 f/s,respectively,which has significant improvement in its comprehensive performance compared with other mainstream object detection algorithms.In the testing under real road conditions,the average warning rate of proposed collision warning system is 96.05%,the missed warning rate and false warning rate is 3.95%and 3.53%,respectively.The warning results of this model are more practical,which can help to improve road traffic safety.关键词
碰撞事故/YOLOv8/车辆碰撞预警/目标检测/碰撞预警区域/防御性驾驶Key words
collision accident/YOLOv8/vehicle collision warning/object detection/collision warning area/defensive driving分类
电子信息工程引用本文复制引用
赵浩帆,宋文爱,章永来,贾媛媛..基于改进YOLOv8的车辆碰撞预警方法研究[J].现代电子技术,2025,48(22):179-186,8.基金项目
山西省重点研发计划基金(202102020101009) （202102020101009）
