Research on rainy day traffic sign recognition algorithm based on PMRNet.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Machine Learning

Background:

• Current traffic sign recognition systems struggle with performance degradation due to rainy weather, where rain obscures targets.
• Existing algorithms often fail to account for the impact of adverse weather conditions like rain on recognition accuracy.

Purpose of the Study:

• To enhance the accuracy of traffic sign recognition in rainy weather conditions.
• To develop a robust algorithm capable of mitigating the effects of rain on image data for intelligent transportation systems.

Main Methods:

• Proposed a two-module algorithm: an image deraining module using Progressive Multi-scale Residual Network (PMRNet) and a traffic sign recognition module using CoT-YOLOv5.
• PMRNet leverages multi-scale residual structures and Convolutional Long-Short Term Memory (ConvLSTM) for effective feature extraction and rain removal.
• CoT-YOLOv5 integrates the Contextual Transformer (CoT) module into YOLOv5 to improve global modeling capabilities and recognition accuracy.

Main Results:

• The PMRNet-based deraining algorithm demonstrated superior performance in removing rain marks, outperforming other methods in Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM).
• The CoT-YOLOv5 algorithm achieved a mean Average Precision (mAP) of 92.1% on the TT100k dataset, a 5% improvement over the original YOLOv5.
• The combined approach significantly enhances traffic sign recognition accuracy under rainy conditions.

Conclusions:

• The proposed PMRNet-based deraining algorithm effectively removes rain artifacts from images.
• The CoT-YOLOv5 algorithm provides a significant improvement in traffic sign recognition accuracy, especially in challenging weather conditions.
• This research offers a promising solution for improving the reliability of intelligent driving and traffic systems in adverse weather.