With the increase of private cars,traffic safety in rainy and foggy days has become an urgent problem to be solved.A voice assisted driving system for driving users in rainy and foggy weather based on machine vision is designed,with the limited embed-ded hardware resources.The system combines the humidity sensors,lightweight dehazing neural network AOD-NET,and object de-tection model YOLOv5n.The K-means++algorithm is used to redesign the anchor frame on the object detection model YOLOv5n.The optimal backbone network is selected to further compress the model size by using the model pruning.The experimental results show that the FPS of the improved model on the Jetson nano reaches by 17.78,and the final mAP by 65.8％on the artificially fogged and resolution changed TT100K(Tsinghua-Tencent 100K)dataset,meeting the practical application of driving assistance in normal weather and rainy and foggy weather.
关键词
机器视觉/Jetson/nano/雨雾天辅助驾驶/交通标志检测/AOD-NET去雾/YOLOv5n
Key words
machine vision/Jetson nano/assisted driving in rain and haze weather/traffic sign detection/AOD-NET dehazing/YOLOv5n
维普
万方数据