Research on the Detection Method of Road Rain, Snow and Icing Based on Image

In view of the high cost and low timeliness of the current road monitoring methods due to rain, snow and icing, this paper proposes an image-based fusion detection method. Through deep learning ResNet50 residual network, a preliminary model for automatic recognition and classification of the images of road weather is obtained, and its average recognition accuracy can reach more than 85%; the icing thickness detection method proposed in this paper can also be measured more accurately. The error range of the icing thickness of the road surface can be controlled within the range of 0.5mm-1mm,