A Hybrid Approach for Lane and Traffic Sign Detection in Foggy Road Conditions

Engineering notes

Key topics: autonomous driving system, autonomous driving, autonomous vehicle, lane detection. See the paper for implementation details and experimental results.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

The reliability of autonomous driving systems, particularly in foggy weather conditions where visibility is limited, is directly dependent on the accurate detection of environmental elements. This study proposes a hybrid method for the simultaneous detection of lane markings and traffic signs in foggy road scenes. In the first stage, a Generative Adversarial Network (GAN)-based model was used to generate 30,000 foggy road images and increase data diversity. Classical image processing methods were compared with a U-Net-based deep learning model for lane detection, while colour and shape-based classical methods were used for traffic sign detection. The findings revealed that U-Net achieved 96% accuracy in lane detection, while classical methods yielded successful results in traffic sign detection with 89% accuracy and low positional error. These results demonstrate that deep learning-based approaches can represent structural complexities more effectively, while classical methods remain a strong alternative for detecting distinct objects. Consequently, a hybrid system combining deep learning for lane detection and classical methods for traffic sign detection offers higher overall accuracy and scene analysis success compared to individual methods, thereby holding the potential to enhance the reliability of autonomous vehicles in foggy weather conditions.

Links and sources

Comments