Advanced Zebra Crosswalk Detection Using Deep Learning Techniques for Smart Transportation Systems

The growing prevalence of road traffic accidents poses significant challenges to vulnerable populations, particularly visually impaired individuals. To enhance safety and accessibility within smart transportation systems, this study presents an advanced detection system for zebra crosswalks, vehicles, and pedestrians utilizing deep learning techniques. The system leverages the Single Shot MultiBox Detector (SSD) model with Transfer Learning for rapid convergence and high accuracy. The dataset, derived from real-world street scenarios, includes nine object classes, enabling the system to provide real-time detection and monitoring. Experimental results demonstrate high precision and recall, underscoring the system's potential to improve road safety and assist traffic management. Future developments will focus on integrating this system into portable devices for broader application.