BROWSE
Detail
Conference Paper
A Study on Automatic Detection of Displacement of Petroglyphs of Bangudae Terrace in Daegok-ri, Ulju Using an Artificial Intelligence Learning Model
Cited 0 time in 
Share 
Share
- Authors
- Sang-Yun Lee, Min-Kyeong Im
- Issue Date
- 2024-11
- Citation
- Future Technologies Conference (FTC) 2024 (LNNS 1154), pp.311-324
- Publisher
- Springer
- Language
- English
- Type
- Conference Paper
- Abstract
- The Petroglyphs of Bangudae Terrace in Daegok-ri, Ulsan are an important human cultural heritage listed on the UNESCO World Heritage Tentative List, but they are being damaged due to encroachment and erosion due to water levels, so continuous monitoring and management are required. In this paper, we propose a method that uses a Deep Learning algorithm to automatically monitor damage and quickly warn when damage occurs. At this time, it is very important to find training data suitable for the model in order to improve the model's performance. In this study, we will present the performance analysis results of two types of data preprocessing, ‘Crop’ and ‘Resize’, based on three models: DeepCrack, PiDiNet, and DexiNed. In Addition, we will compare and analyze the displacement detection performance of Deep Learning models according to three labeling methods, including ‘Linestrip’, ‘Polygon’, and the case of applying ‘Linestrip’ and ‘Polygon’ simultaneously. And, we will present experimental results to determine which labeling method is more suitable for which model. DeepCrack showed good results in detecting cavities when trained with data containing ‘Polygon’. And DexiNed was good at detecting joint separations when trained with data including ‘Linestrip’. PiDiNet was good at detecting joint separations and cavities regardless of the labeling method. As a result, we found that the ‘Linestrip’ method is a better labeling method when detecting the boundary area of displacement, and the ‘Polygon’ method is a better labeling method when detecting the displacement area.
- KSP Keywords
- Automatic Detection, Continuous monitoring, Cultural Heritage, Data Preprocessing, Labeling method, Monitoring and management, Performance analysis, UNESCO world heritage, Water level, artificial intelligence, deep learning(DL)