Foggy day traffic sign detection method based on deep learning

The invention relates to the technical field of image recognition target detection, in particular to a foggy day traffic sign detection method based on deep learning, and the method comprises the steps: S1, constructing a traffic sign detection model; the traffic sign detection model is obtained based on YOLOv5 improvement; the improvement comprises the following steps: increasing shallow layer features and sampling depth at a neck NECK through cascade operation, adding a corresponding small target detection head at a head HEAD, and adjusting three feature maps P2, P3 and P4 output by an original feature pyramid network FPN into P1, P2, P3 and P4; a multi-dimensional dynamic detection head SSTHead for a small target is added between the neck NECK and the head HEAD, and the multi-dimensional dynamic detection head SSTHead is used for carrying out adaptive learning on different attention maps in different dimensions and filtering the inconsistency between the features of each dimension; s2, acquiring a training