DehazeNet-assisted fog removal and channel estimation for robust visible light V2V communication

In foggy weather conditions, road visibility is drastically reduced, significantly increasing the risk of head-on vehicle collisions. This paper proposes a vehicle-to-vehicle (V2V) communication system based on visible light communication (VLC), enhanced by a deep neural network (DNN) that estimates and corrects the channel degradation caused by fog. Traditional channel models based on physical parameters often fail under highly dynamic foggy scenarios. To address this limitation, we introduce a DehazeNet-based framework that leverages real-time camera input to extract fog characteristics and estimate the attenuation effects on the VLC channel. DehazeNet, originally designed for single image haze removal, is adapted to model and correct signal path loss due to fog-induced scattering and absorption. This enables real-time compensation of visual and communication degradation, improving both signal integrity and object visibility for vehicle detection. A time-to-collision (TTC) metric is computed from the estimated inter-vehicle distances to trigger timely warnings. At a moderate fog visibility level of 50 m, the original system shows a TTC of 1.08 s at 42 m, while the fog-suppressed system improves this to 1.40 s at 54.47 m. Thus, simulation results demonstrate that the DehazeNet-based compensation significantly enhances the effective range and accuracy of V2V communication in foggy environments, enabling earlier collision warnings and contributing to improved road safety.