Development of an AI-Driven Platform for Autonomous Dashcam Data Integration with the International Road Assessment Program Visual Data Analysis for Road Safety Assessment Enhancement

Abstract

Abstract. This study develops an AI-driven platform to autonomously convert unstructured dashcam footage into structured data compatible with the Visual iRAP Data Analysis (VIDA) traffic management system. The platform utilizes YOLOv7 for object detection and vehicle trajectory extraction, Natural Language Processing (NLP) for text and audio analysis, and predictive models to clean and structure metadata into VIDA-compatible formats (CSV, XML). Tested on 100 dashcam videos from Saudi Arabian roads—including highways, urban intersections, and accident zones—the platform achieved a 96% metadata extraction success rate with processing times as low as 3.8 seconds per video. Usability testing yielded a 4.5/5 satisfaction rating from urban planning professionals. This work bridges the critical gap between raw vehicular data and intelligent transportation systems, enabling data-driven road safety assessment and supporting Saudi Vision 2030's smart city objectives.