Advancing Driver Assistance Systems: Multi-Class Weather Classification using Supervised Learning for Enhanced Traffic Safety

Abstract

Traffic accidents present considerable hazards in adverse weather conditions, including rain, nighttime, icy surfaces, and locations with insufficient street lighting. Modern driver assistance systems are predominantly engineered for functionality in optimal weather conditions. The classification of weather conditions, which entails the recognition of optical characteristics within visual data, represents a viable method to enhance computer vision capabilities in adverse weather scenarios. This study presents a multi-class weather classification system that utilizes various weather features and supervised learning techniques to improve the effectiveness of computer vision in challenging weather conditions. The procedure involves extracting essential visual characteristics from various traffic images, which are subsequently converted into an eight-dimensional feature space. Five distinct supervised learning techniques are subsequently employed for model training. The analysis of the extracted features indicates that the proposed approach enhances the accuracy and adaptability of weather classification, establishing a foundation for improvements in vehicular safety. This innovation enhances its functionality in scenarios characterized by low visibility at night and on icy surfaces, thereby improving the driver's field of vision. Feature extraction serves as a vital component in pattern recognition, providing an effective approach to streamline high-dimensional image data and isolate key information necessary for interpreting multi-traffic scenarios.