ANX-Net: A Fast and Resource Optimized Network for Image Dehazing for Driving in Haze Weather Conditions

Severe weather conditions such as haze will bring serious problems to the safe driving of autonomous vehicle. In order to ensure that autonomous vehicle can still run safely in frequent bad weather, the research of image dehazing algorithm is very important. The key to safe and reliable driving is that autonomous vehicle can obtain clear images in severe haze weather conditions. Therefore, ensuring the dehazing performance of the dehazing algorithm is very important. In this paper, we propose ANX-Net, which is a robust and reliable dehazing network for autonomous vehicle. The network uses components such as feature extraction module, channel attention module, multi-scale spatial attention module and gsconv module to effectively dehaze the images taken by the autonomous vehicle camera. Through a detailed qualitative and quantitative evaluation of the road traffic dataset AAR in hazy weather, the effectiveness of the proposed network was analyzed, demonstrating its good dehazing performance.