Damage Assessment of Earthen Sites of the Ming Great Wall in Qinghai Province
Yumin Du, Wenwu Chen, Kai Cui, Jingke Zhang, Zhuo Chen, Qiyong Zhang
Abstract
Being a world-renowned heritage site, the Ming Great Wall, located in Qinghai Province, China, mainly has existed as the form of earthen sites with high historical, artistic, and scientific values. However, exposed under environmental impacts for about 500 years, these sites have been seriously threatened. The study on their damage assessment using reasonable methods is the key premise for further protection work. As there are few studies focused on the damage assessment of earthen sites, especially by using machine learning methods, this study explored to apply the current effective machine learning approaches, namely Support Vector Machine (SVM) and back-propagation (BP) Neural Network, into solving this problem. The authors used two such algorithms by training and testing the existing data of damage assessment results of 18 earthen sites located in Qinghai Province. By comparing experimental results, the prediction effects of SVM are much better than those of BP Neural Network. To test its practicability, Liutun Great Wall was chosen as an engineering case, showing better prediction and reasonable performance of SVM. This research has proved that SVM can be selected as a suitable model to assess the damage levels of earthen sites located in Qinghai Province for their future conservation works.