Robust Lane Detection with Wavelet-Enhanced Context Modeling and Adaptive Sampling

2025-03-24

AI Summary

This paper addresses the insufficient robustness of lane detection under challenging conditions—including adverse weather, abrupt illumination changes, severe occlusion, and complex curved roads—by proposing the Wavelet-Enhanced Feature Pyramid Network (WE-FPN). Its key contributions are: (1) a novel wavelet-driven non-local context modeling module that enhances long-range structural representation; and (2) an illumination-adaptive preprocessing strategy coupled with a distance- and curvature-aware attention-guided sampling mechanism, improving localization accuracy for distant curved lanes and occluded regions. Evaluated on CULane and TuSimple benchmarks, WE-FPN outperforms state-of-the-art methods such as CLRNet: it achieves a 3.2% improvement in mean F1-score under rain/fog, low-light, and severe occlusion conditions, and a 5.7% gain in detection accuracy for distant curved lanes. These results demonstrate WE-FPN’s strong generalization capability and practical effectiveness for real-world lane detection.

Technology Category

Computer Vision: Learning & Optimization for CVNatural Language Processing: Safety and RobustnessMachine Learning: Multimodal Learning

Application Category

Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphsSearch and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingResponsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Lane detection is critical for autonomous driving and ad-vanced driver assistance systems (ADAS). While recent methods like CLRNet achieve strong performance, they struggle under adverse con-ditions such as extreme weather, illumination changes, occlusions, and complex curves. We propose a Wavelet-Enhanced Feature Pyramid Net-work (WE-FPN) to address these challenges. A wavelet-based non-local block is integrated before the feature pyramid to improve global context modeling, especially for occluded and curved lanes. Additionally, we de-sign an adaptive preprocessing module to enhance lane visibility under poor lighting. An attention-guided sampling strategy further reffnes spa-tial features, boosting accuracy on distant and curved lanes. Experiments on CULane and TuSimple demonstrate that our approach signiffcantly outperforms baselines in challenging scenarios, achieving better robust-ness and accuracy in real-world driving conditions.

Problem

• Improves lane detection in adverse conditions like weather and occlusion
• Enhances global context modeling for occluded and curved lanes
• Boosts accuracy on distant and curved lanes with adaptive sampling

Innovation

• Wavelet-Enhanced Feature Pyramid Network for context modeling
• Adaptive preprocessing module for poor lighting conditions
• Attention-guided sampling strategy for spatial features