Research on Methods of Expressway Vehicle Detection under Abnormal Weather Conditions Based on Deep Learning

Aiming at the issue that vehicle detection accuracy is easily influenced by abnormal weather conditions such as rain, snow, and frog, et al. This paper studies the methods of expressway vehicle detection based on deep learning. First, methods for detecting expressway vehicles based on Faster-RCNN, YOLOV3, and SSD are compared and analyzed. Then, based on SEU expressway vehicle detection dataset under abnormal weather conditions, the training of the vehicle detection model and the research of compare experiments on Faster-RCNN, YOLOv3, and SSD are carried out by manually labeling and collecting specific regions of vehicles. Theoretical analysis and experimental results show that the YOLOv3-based detection model of expressway vehicle detection under abnormal weather conditions outperforms the other two methods, with an average accuracy of 99.2%.