Experiment and Analysis of Tilt Prediction for Anguk-dong Byeolgung Roof based on EfficientNet according to Integration by Environment

doi:None

All Issue

2023 Vol.8, Issue 3 Preview Page Next Page

Article

Experiment and Analysis of Tilt Prediction for Anguk-dong Byeolgung Roof based on EfficientNet according to Integration by Environment 환경별 통합에 따른 EfficientNet 기반 안국동 별궁 지붕부 기울음 예측 실험 및 분석: 이상윤^1†ㆍ안정서²
Lee, Sang-Yun^1†ㆍAn, Jeongseo²; ¹한국전자통신연구원 디지털융합연구소 과학치안공공ICT연구센터 책임연구원
²공주대학교 문화재보존과학과 학사과정 3학년

¹Principal Researcher, Police Science & Public Safety ICT Research Center, Digital Convergence Research Laboratory, Electronics and Telecommunications Research Institute
²BS, Cultural Heritage Conservation Science Kongju University

31 December 2023. pp. 197-210

PDF

Abstract

Wooden cultural heritages can be deteriorated due to the aged materials or natural disasters such as localized heavy rains. Regular monitoring is necessary to accurately identify displacements and take precautions against anticipated problems. However, regular monitoring has limitations in terms of time and cost. So, If Deep Learning technology is used when displacement occurs, it will be a great help to the conservation of cultural heritages. In this study, we constructed data for each environment using CCTV data. Then, we confirmed whether it is possible to predict the tilt of the roof using EfficientNetB0 model. As a result of experiment, it showed prediction accuracy of 99.68% or higher in all environments such as clear, overcast, cloud, snow, and rain. Based on the results of the model verification, we analyzed the combinations that demonstrate high predictive accuracy when integrating environments. Integrated datasets were constructed for both environments, and experiments were conducted, showing high prediction accuracy of over 99.84% for both the same and different variations in all environments. We propose optimal prediction models for tilt of roof validating their performance on 14 datasets that were not used to train the respective models.

Keywords

Wooden cultural heritages

Monitoring

Deep learning

Artificial intelligence

Anguk-dong byeolgung

목조문화재는 나무의 재료의 특성상 노후화가 진행되어 다양한 변위가 발생하기도 하며, 집중호우와 같은 자연재해로 인해 문화재에 변위가 나타나기도 한다. 변위들을 정확히 파악하고 예상되는 문제에 대한 예방책을 강구하기 위해서는 정기적인 모니터링이 요구된다. 하지만 정기적인 모니터링은 시간적, 비용적 측면에서 한계가 존재하기 때문에 딥러닝을 활용한 모니터링 방식이 필요하다. 본 연구에서는 CCTV 영상 데이터를 활용하여 맑음, 흐림, 구름, 눈, 비 환경에 대한 데이터셋을 구축한 후 EfficientNetB0 모델로 실험을 진행했다. 실험 결과 모든 환경이 동일한 변형과 다양한 변형에서 99.68% 이상의 예측 정확도를 보여주었다. 모델 성능 검증 결과를 통해 환경을 통합할 때 높은 예측 정확도를 보여줄 수 있는 조합을 분석하였다. 이를 바탕으로 2가지 환경에 대한 통합 데이터셋을 구축해 2가지 환경통합 실험을 한 결과, 모든 환경에서 동일한 변형과 다양한 변형이 99.84% 이상의 높은 예측 정확도를 보여주었다. 생성한 모델마다 학습에 사용하지 않은 14개의 데이터셋으로 모델 성능 검증을 통해 최적의 지붕부 기울음 예측 모델을 제안한다.

키워드

목조 문화재

모니터링

딥러닝

인공지능

안국동 별궁

References

An, J.S., Lee, S.Y. (2023). “Prediction of Roof Displacement for Anguk-dong byeolgung using EfficientNet,” IEIE, pp.1046-1049.
Choi, W.W., Park, J.Y., Lee, S.Y. (2021). “Detection of Abnormalities in Wooden Cultural Properties Using the F-anogan Model,” IEIE, pp.168-169.
Cultural Heritage Administration (2009). Restoration and Repair Report on the Relocation of Anguk-dong byeolgung. Cultural Heritage Administration, Daejeon, Korea.
Cultural Heritage Administration (2020). The First Master Plan for Preservation, Management, and Utilization R&D of cultural Heritage (2021~2025). Cultural Heritage Administration, Daejeon, Korea.
Cultural Heritage Administration (2022). Basic Manual of Cultural Heritage Disaster Safety Management. Cultural Heritage Administration, Daejeon, Korea.
Jo, S.S. (2019). “Trends and Direction of State-run Researches on the Disaster Prevention of Cultural Heritages,” Journal of the Society of Cultural Heritage Disaster Prevention, Vol.4, No.2, pp.103-110.
Kang, J.Y., Kim, J.Y., Lim, H.S., Kwak, J.H. (2021). “A Crack Detection of Wooden Cultural Assets using EfficientNet model,” KSCI, pp.125-127.
Kang, S.J., Lee, S.Y. (2023). “A study on Performance Improvement of Heunginjimun Roof Displacement Determination Model Using Artificial Intelligence Algorithm,” IEIE, pp.1050-1054.
Kim, S.D. (2020). “A Study on the Practice and Case Study of Managing Architectural Cultural Heritage of CHA in Relation to the Damage Prevention of State-designated Cultural Heritage and State-registered Cultural Heritage - Focused on Regular Surveys, Safety Inspections, Ambient Vibration Measurements, and Disaster Impact Surveys,” Journal of the Society of Cultural Heritage Disaster Prevention, Vol.5 No.2, pp.73-88.
Kim, S.H., Jo, S.S., Lee, S.H., Lee, H.S. (2021). “Analysis of Damage to Wooden Architectural Heritage Caused by the 9.12 Earthquake and the Pohang Earthquake,” Journal of The Society of Cultural Heritage Disaster Prevention, Vol.6, No.2, pp.105-119.
Lee, S.Y., Park, J.Y. (2022). “Ensemble Deep Learning Model for Prediction of Tilt Displacement for Immovable Cultural Heritage,” IEIE, pp.1397-1398.
Moon, H.C. (2018). “A Study on the Improvement of Cultural Heritage Protection System in the Process of Fourth Industrial Revolution,” Vol.3, No.2, pp.129-139.
Park, J.Y., Lee, S.Y. (2022). “Displacement Detection of Wooden Cultural Properties Using Unsupervised Learning,” ICCE-Asia, pp.302-305. 10.1109/ICCE-Asia57006.2022.9954670
Selvaraju, R.R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., Batra, D. (2017). “Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization,” IEEE, pp.618-626. 10.1109/ICCV.2017.74
Tan, M.X., Quoc, L. (2019). “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks,” PMLR, pp.6105-6114.

Information

Publisher :The Society of Cultural Heritage Disaster Prevention
Publisher(Ko) :문화재방재학회
Journal Title :Journal of the Society of Cultural Heritage Disaster Prevention
Journal Title(Ko) :문화재방재학회논문집
Volume : 8
No :3
Pages :197-210

[1] An, J.S., Lee, S.Y. (2023). “Prediction of Roof Displacement for Anguk-dong byeolgung using EfficientNet,” IEIE, pp.1046-1049.

[2] Choi, W.W., Park, J.Y., Lee, S.Y. (2021). “Detection of Abnormalities in Wooden Cultural Properties Using the F-anogan Model,” IEIE, pp.168-169.

[3] Cultural Heritage Administration (2009). Restoration and Repair Report on the Relocation of Anguk-dong byeolgung. Cultural Heritage Administration, Daejeon, Korea.

[4] Cultural Heritage Administration (2020). The First Master Plan for Preservation, Management, and Utilization R&D of cultural Heritage (2021~2025). Cultural Heritage Administration, Daejeon, Korea.

[5] Cultural Heritage Administration (2022). Basic Manual of Cultural Heritage Disaster Safety Management. Cultural Heritage Administration, Daejeon, Korea.

[6] Jo, S.S. (2019). “Trends and Direction of State-run Researches on the Disaster Prevention of Cultural Heritages,” Journal of the Society of Cultural Heritage Disaster Prevention, Vol.4, No.2, pp.103-110.

[7] Kang, J.Y., Kim, J.Y., Lim, H.S., Kwak, J.H. (2021). “A Crack Detection of Wooden Cultural Assets using EfficientNet model,” KSCI, pp.125-127.

[8] Kang, S.J., Lee, S.Y. (2023). “A study on Performance Improvement of Heunginjimun Roof Displacement Determination Model Using Artificial Intelligence Algorithm,” IEIE, pp.1050-1054.

[9] Kim, S.D. (2020). “A Study on the Practice and Case Study of Managing Architectural Cultural Heritage of CHA in Relation to the Damage Prevention of State-designated Cultural Heritage and State-registered Cultural Heritage - Focused on Regular Surveys, Safety Inspections, Ambient Vibration Measurements, and Disaster Impact Surveys,” Journal of the Society of Cultural Heritage Disaster Prevention, Vol.5 No.2, pp.73-88.

[10] Kim, S.H., Jo, S.S., Lee, S.H., Lee, H.S. (2021). “Analysis of Damage to Wooden Architectural Heritage Caused by the 9.12 Earthquake and the Pohang Earthquake,” Journal of The Society of Cultural Heritage Disaster Prevention, Vol.6, No.2, pp.105-119.

[11] Lee, S.Y., Park, J.Y. (2022). “Ensemble Deep Learning Model for Prediction of Tilt Displacement for Immovable Cultural Heritage,” IEIE, pp.1397-1398.

[12] Moon, H.C. (2018). “A Study on the Improvement of Cultural Heritage Protection System in the Process of Fourth Industrial Revolution,” Vol.3, No.2, pp.129-139.

[13] Park, J.Y., Lee, S.Y. (2022). “Displacement Detection of Wooden Cultural Properties Using Unsupervised Learning,” ICCE-Asia, pp.302-305. 10.1109/ICCE-Asia57006.2022.9954670

[14] Selvaraju, R.R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., Batra, D. (2017). “Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization,” IEEE, pp.618-626. 10.1109/ICCV.2017.74

[15] Tan, M.X., Quoc, L. (2019). “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks,” PMLR, pp.6105-6114.

Journal of the Society of Cultural Heritage Disaster Prevention ISSN:2508-2825(Print) 문화재방재학회논문집

All Issue