DMRVisNet: Deep Multihead Regression Network for Pixel-Wise Visibility Estimation Under Foggy Weather

Scene perception is essential for driving decision-making and traffic safety. However, fog, as a kind of common weather, frequently appears in the real world, especially mountain areas, making it difficult to accurately observe surrounding environments. Therefore, precisely estimating visibility under foggy weather can significantly benefit management To address this, most current methods use professional instruments outfitted at fixed locations on roads perform measurement; these are expensive less flexible. In this paper, we propose an innovative end-to-end convolutional neural network framework estimate leveraging Koschmieder’s law image data. The proposed method estimates by integrating physical model into framework, instead directly predicting value via network. Moreover, pixel-wise map against those previous measurement which solely predict single entire image. Thus, estimated result our more informative, particularly uneven fog scenarios, developing precise early warning system thereby better protecting intelligent transportation infrastructure systems promoting their development. validate virtual dataset, FACI, containing 3,000 images different concentrations, collected using AirSim platform, available https://github.com/coutyou/FoggyAirsimCityImages . Detailed experiments show that achieves performance competitive state-of-the-art methods.