基于YOLOv3的前车行驶状态识别与预警技术研究

YOLOv3-Based Driving Status Recognition and Warning System

在交通事故中接近60%是由于车与车的追尾碰撞所造成。更有大量研究显示,如果能在车辆追尾事故发生前一秒给驾驶员发出预警,可以避免90%以上的交通事故发生。因此,为了预防追尾事故的发生,开展对前车行驶状态识别与预警的研究很有必要。使用YOLOv3算法来深度学习训练并生成行驶中的车辆尾部检测模型,并将此模型作为车尾灯状态识别的基础框架。设计了车尾灯状态识别模型与语音报警系统,构成一个前车行驶状态识别与预警的模型。通过实验得出模型对前车状态识别的准确率达87.5%,即设计的前车行驶状态识别与预警模型具有一定的可行性。

Nearly 60%of traffic accidents are caused by rear-end collisions between vehicles.A large number of studies have shown that if the driver could be alerted one second before a rear-end collision occurs,90%of traffic accidents could be avoided.Therefore,in order to prevent the occurrence of rear-end collision accidents,it is necessary to carry out research on the identification and early warning of the driving state of the preceding vehicle.The YOLOv3 algorithm was used to train and generate a moving vehicle tail detection model by deep learning,and the model was used as the basic framework for taillight state recognition.The taillight status recognition model and voice alarm system were designed to form a model for recognition and early warning of the driving state of the preceding vehicle.Through experiments,it was concluded that the accuracy of the model for the recognition of the preceding vehicle state was 87.5%,that was,the designed preceding vehicle driving state recognition and early warning model had certain feasibility.