본문 바로가기
  • Home

A Study on Automated Enforcement Decision Methods for Urban Advertising Materials Using Vision-Language Models and Evidence-Based Reasoning

  • Journal of Internet of Things and Convergence
  • Abbr : JKIOTS
  • 2025, 11(6), 25
  • Publisher : The Korea Internet of Things Society
  • Research Area : Engineering > Computer Science > Internet Information Processing
  • Received : December 3, 2025
  • Accepted : December 22, 2025
  • Published : December 31, 2025

박지우 1 Kang taewook 1 Chang-Sun Shin 1 Cho Yong Yun 1 PARK CHUL YOUNG 1

1국립순천대학교

Accredited

ABSTRACT

This study proposes an automatic classification method combining a Vision-Language Model (VLM) with evidence-based reasoning to efficiently identify illegal banners in urban environments. The system analyzes the type and textual content of advertisement images and enhances decision reliability through Rule-of-2 and Verifier procedures. Experimental results show an overall accuracy of approximately 70%, with F1-scores of 0.73 for the illegal class and 0.65 for the legal class. Confusion matrix analysis indicates 8 false positives and 24 false negatives. These findings suggest the need to reduce false positives while maintaining illegal detection performance. This study demonstrates the potential for automating urban advertisement enforcement and highlights the need for extensive datasets and integration with administrative systems for rigorous validation.

Citation status

* References for papers published after 2024 are currently being built.