Quantitative Analysis of Effective Road Width for Fire Engine Access under On-Street Parking Using Drone Imagery and Deep Learning

This study proposes a method to quantitatively estimate the effective road width available for fire engine access in on-street parking environments using drone imagery and deep learning-based object detection. Vehicle objects were detected using a YOLOv8 model, and occupancy areas were delineated through Canny edge-based boundary extraction. The minimum passage width was estimated via a y-coordinate iterative search and converted into real-world units using the Ground Sample Distance (GSD) derived from standardized lane markings. Experimental results showed that effective road widths ranged from approximately 2.8 m to 3.8 m, indicating that on-street parking can significantly constrain fire engine accessibility regardless of the road’s planned geometric layout. The proposed approach provides a quantitative image-based methodology for identifying urban fire safety vulnerabilities and supporting evidence-based emergency response planning.