<![CDATA[Image-based weather recognition has become an important research area due to its applications in intelligent transportation systems, autonomous vehicles, road condition monitoring, and computer vision. Recent advances in deep learning have significantly improved the automatic recognition of diverse weather conditions. However, the variety of image sources, datasets, deep learning architectures, and weather scenarios makes it challenging to obtain a comprehensive understanding of current research trend. Therefore, this study aims to identify the image sources and datasets used in weather recognition research, analyze dominant deep learning approaches, and investigate the weather conditions most frequently addressed in image-based weather recognition and road weather detection. A Systematic Literature Review (SLR) following the PRISMA 2020 guidelines was conducted using the Scopus database. From an initial set of 603 retrieved articles, 73 studies met the predefined inclusion criteria and were selected for further analysis. The results indicate that public datasets such as DAWN, RTTS, Foggy Cityscapes, nuScenes, WeatherDataset-4, and WeatherNet are among the most frequently used data sources. Convolutional Neural Networks (CNNs) remain the dominant approach, although the adoption of Transformer-based models, Vision Transformers, YOLO, Multimodal Fusion, and Multi-Task Learning has increased considerably in recent years. Furthermore, rain, fog, and snow are the most extensively investigated weather conditions due to their significant impact on visibility and perception system performance. The findings provide a comprehensive overview of recent developments in deep learning-based weather recognition and offer valuable insights and directions for future research.]]>
Copyrights © 2026
