Research on Traffic Sign Image Recognition Algorithms Under Complex Weather Conditions

In the transportation system, the influence of haze is more significant, such as license plate recognition, real-time monitoring, etc. The visibility of both people and equipment is greatly affected in foggy weather, leading to the emergence of foggy image processing. We analyzed the recognition requirements of traffic signs in foggy weather and conducted research on algorithms for removing fog from foggy images and extracting image edges. This topic mainly improved on the traditional Retinex algorithm, recognizing the loss of detail information in images under Gaussian filtering conditions. We applied guided filtering to the estimation of illumination images to achieve the preservation of image edge information. In terms of image recognition, the currently best performing LOG operator and Canny edge extraction algorithm were applied to achieve the extraction of detail information. Then, based on the background knowledge of Convolutional neural network, a small Convolutional neural network model is designed for training to realize the recognition and classification of traffic sign images. The experimental results show that the method proposed in this paper can achieve good functions in fog removal and traffic sign recognition.