SCOUT: A SOLUTION TO AUTOMATIC INSPECTION OF THE OUTDOOR ADVERTISEMENT BANNERS

This paper presents the design and implementation of SCOUT - Street-level Capture & Overlay for Urban Track-inspection - a hybrid AI system developed for the automated detection and geospatial mapping of outdoor advertisement banners in urban environments. Targeting the operational needs of municipal authorities and marketing service agencies, SCOUT integrates rule-based heuristics, large language models (LLMs), and machine learning (ML) techniques, including a custom-trained YOLOv11 object detector. The system processes dashcam video streams to identify advertisement banners and automatically logs their locations in a spatial database using GPS metadata. Comparative experiments evaluate the performance of traditional computer vision methods, LLM-based analysis, and deep learning models, highlighting the limitations of classical and LLM-based approaches in dynamic urban settings. Results demonstrate that the hybrid system - anchored by a lightweight object detection model - offers significant improvements in detection accuracy, scalability, and real-time applicability. SCOUT provides a novel, practical solution for urban advertisement monitoring and policy enforcement, with potential for integration into larger smart city infrastructure through third-party APIs.