Spatio-Temporal Transportation Images Classification Based on Light and Weather Conditions

Abstract

Advancements transportation systems and proliferation of imaging technologies have provided an safety and efficiency through intelligent analysis using spatio-temporal images. This research focuses on the development of a robust classification system capable of discerning transportation images based on both spatial and temporal features, with a specific emphasis on light and weather conditions. The study begins with the collection of a diverse dataset encompassing various lighting scenarios (day, night, dawn, dusk) and weather conditions (sunny, rainy, snowy, foggy). Through meticulous preprocessing, images are standardized and relevant metadata, including timestamps and location information, is extracted. Feature extraction techniques, such as color histograms and texture features, are employed to capture spatial characteristics, while pre-trained Enchanced convolutional neural networks (ECNNs) aid in learning high-level spatial representations. To account for temporal dependencies in transportation images, a Hybrid novel neural network architecture is designed, incorporating recurrent neural networks (RNNs) or 3D CNNs or ECNN. The model is trained on the labeled dataset, enabling it to predict classes associated with different lighting and weather conditions through the implementation of softmax layers in the output. Evaluation of model is identifying based on accuracy, precision, recall, and F1 score. The satisfactory results i.e. more than 90% accuracy performance received using proposed model and it is useful to deploy for real-time or batch classification of spatio-temporal transportation images. Continuous improvement is emphasized through regular updates with new data to adapt to evolving lighting and weather conditions. The proposed classification system presents a promising avenue for enhancing transportation safety and efficiency through intelligent image analysis, with potential applications in autonomous vehicles, traffic management, and emergency response systems.