Innovative Multi-Feature Based Weather Classification for Supervised Learning in Multiclass Environments

Abstract

Highway traffic accidents have devastating consequences, leading to significant loss of lives and property. However, one promising solution to reduce these accidents lies in the implementation of advanced driver assistance systems (ADAS). These systems have proven to be effective in enhancing road safety. A critical component for these ADAS is the ability to perceive and understand complex traffic scenes under various weather conditions, as this valuable information can greatly improve their performance. Different weather conditions present unique challenges, particularly in terms of visibility, and specialized approaches are needed to address these challenges effectively. By tailoring the ADAS algorithms based on weather categories, we can enhance visibility and expand the application of these systems further. Among the weather conditions that significantly impact traffic safety are rainy days, dark nights, overcast and rainy nights, foggy days, and other situations with poor visibility. Most current vision-based driver assistance systems are optimized to function well under favorable weather conditions. To address the issue of poor visibility in bad weather situations, a multi-class weather classification method is proposed. This method relies on multiple weather features and supervised learning techniques. First, visual features are extracted from multi-traffic scene images. These features are then represented as an eight-dimensional feature matrix, capturing the crucial characteristics of the scene. Next, the supervised learning algorithms are employed to train classifiers, enabling the system to recognize different weather conditions accurately. The analysis of the proposed method shows promising results. The extracted features effectively describe the image semantics, and the trained classifiers demonstrate high recognition accuracy and adaptability. This lays the foundation for enhancing anterior vehicle detection during nighttime illumination changes and improving the driver's field of vision on foggy days. By integrating this multi-class weather classification method into ADAS, there will be a significant improvement in road safety, especially during challenging weather conditions. This advancement will have a positive impact on reducing traffic accidents and protecting the lives of motorists and pedestrians alike