Anomaly Detection in Masonry Minarets Through Hybrid Video-Based Deep Learning Methodologies

HACIEFENDİOĞLU K., KAHYA V., Aslan T., Duman C., BAŞAĞA H. B., OKUR F. Y., ...Daha Fazla

International Journal of Architectural Heritage, 2025 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Basım Tarihi: 2025
  • Doi Numarası: 10.1080/15583058.2025.2524764
  • Dergi Adı: International Journal of Architectural Heritage
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Arts and Humanities Citation Index (AHCI), Scopus, Aerospace Database, Art Source, Communication Abstracts, Compendex, Geobase, Index Islamicus, Metadex, Civil Engineering Abstracts
  • Anahtar Kelimeler: Anomaly detection, autoencoder, convolutional neural network, long short-term memory, masonry minarets, optical flow, structural health monitoring
  • Karadeniz Teknik Üniversitesi Adresli: Evet

Özet

This study proposes a novel video-based structural health monitoring methodology tailored specifically for historic masonry minarets. It integrates advanced optical flow image processing and deep learning methods (Autoencoder, CNN, and LSTM) for the non-contact detection of structural anomalies. A scaled minaret model was experimentally monitored using a multi-camera system under controlled damage scenarios. Optical flow analysis provided precise displacement measurements, which were analyzed with a hybrid Autoencoder- PCA model to enhance the accuracy of anomaly detection. The results indicate that Autoencoder and CNN models effectively identified structural anomalies, demonstrating consistent and reliable performance. Although the LSTM model captured temporal dynamics well, it exhibited higher false-positive rates, indicating room for improvement. Overall, the methodology presents a reliable and efficient non-contact tool for preserving and monitoring the structural integrity of historic masonry minarets.