Traffic Sign Detection Based on Improved YOLOv3 in Foggy Environment

Abstract

Aiming at the problem of poor detection accuracy and inaccurate positioning of traffic signs under foggy conditions, this paper proposes an improved YOLOv3 detection algorithm. Firstly, a data set of Chinese traffic signs in a foggy environment is constructed; The dark channel a priori algorithm based on guided filtering is used to process the image with fog, which overcomes the problem of image quality degradation caused by fog. Mosaic data enhancement is performed on the annotated data set image, which speeds up the convergence speed of the network. Increased the feature scale of YOLOv3 algorithm. The loss function of the network is optimized, CIOU is used as the positioning loss, and the positioning accuracy is improved. At the same time, the method of transfer learning is used to overcome the problem of insufficient samples. The enhanced yolov3 algorithm proposed in this paper has higher detection accuracy and shorter detection time than the standard yolov3 algorithm and SSD algorithm.