Lane Departure Accident Prevention in Foggy Conditions: A Prior-Guided Dynamic Feature Fusion Transformer Framework for Real-Time Lane Detection

TL;DR

This work tackles lane departure risk in foggy conditions by introducing PDT-Net, a prior-guided dynamic feature fusion transformer for real-time lane detection. The framework combines a Global Feature Fusion Module to preserve global context, a Dynamic Feature Fusion Module to model relationships among lane instances, and a Prior-Guided Edge Enhancement Module to recover edge details lost in fog, all built on a transformer-based backbone. It introduces the FoggyLane dataset and foggy variants of CULane and Tusimple to address data gaps, and demonstrates state-of-the-art performance across all foggy benchmarks with real-time inference on edge hardware, underscoring practical impact for ADAS safety. The results show significant gains from each module, particularly the global fusion component, and validate the approach's applicability to real-world driving scenarios where visibility is compromised.

Abstract

Lane departure accident prevention plays a critical role in enhancing road safety, and lane detection is a core technology to achieve this goal, especially under complex weather conditions. While existing lane detection algorithms perform well under favorable weather conditions, their effectiveness significantly degrades in foggy environments, which increases the risk of traffic accidents. In response to this challenge, we propose PDT-Net, a robust Prior-Guided Dynamic Feature Fusion Transformer framework designed for real-time lane detection in foggy conditions. This framework integrates three key modules: a Global Feature Fusion Module (GFFM) to capture the relationship between local and global features in foggy images, a Dynamic Feature Fusion Module (DFFM) to model the structural and positional relationships of lane instances, and a Prior-Guided Edge Enhancement Module (PEM) to recover lost edge details in foggy environments. Furthermore, we introduce the FoggyLane dataset, a real-world dataset that specifically targets lane detection in foggy conditions, along with two synthesized datasets, FoggyCULane and FoggyTusimple, to address the lack of fog-specific data for lane detection. Extensive experiments show that PDT-Net achieves state-of-the-art performance with F1-scores of 95.04% on FoggyLane, 79.85% on FoggyCULane, and 96.95% on FoggyTusimple. Moreover, with TensorRT acceleration, our method achieves a processing speed of 38.4 FPS on the NVIDIA Jetson AGX Orin, confirming its real-time capability and robustness in challenging foggy environments. By improving the precision of lane detection, our framework can contribute to active safety warning systems, helping to prevent accidents in foggy conditions.

Figures (15)

• Figure 1: Lane detection in foggy scenarios for ADAS.Pexels Inaccurate lane detection may lead to lane departure, increasing the risk of collision. Accurate lane detection provides essential technical support for the active safety functions of vehicles. Once the vehicle deviates from its intended lane, the ADAS can promptly issue an active safety warning to alert the driver, enabling timely corrective actions and ultimately preventing the occurrence of traffic accidents.
• Figure 2: Overview of the proposed network.
• Figure 3: Backbone Network with Global Feature Fusion Module(GFFM).
• Figure 4: The structure of the Dynamic Feature Fusion Module(DFFM)
• Figure 5: Comparison of edge information between foggy and normal scenarios.