THz spectra-based SVM prediction model for Yungang Grottoes samples. (March 2015)

• Record Type: Journal Article
• Authors: Meng, Tianhua; Du, Ruiqing; Hou, Zheng; Yang, Jin; Zhao, Guozhong
• Is Part Of: Journal of archaeological science. Volume 55(2015:Mar.)
• Page Start: 280
• Page End: 285
• Publication Date: 2015-03
• Subjects: THz-TDS -- SVM -- Transmission coefficient -- Prediction model -- Stone cultural relics; Archaeology -- Periodicals; Archéologie -- Périodiques; 930.1
• DOI: 10.1016/j.jas.2015.01.012
• Languages: English
• ISSNs: 0305-4403

Abstract

Abstract: Knowledge of weathered depth is critically important for repairing and protecting stone relics that are open to elements. Weathered depth is an important basis of stability/durability evaluation and protection plan formulation for stone carving and grottoes. Weathered depth is the basis for identifying the part that has been weathered or the material properties that have been changed, as well as the basis for determining weathered speed or weathered depth. Weathered depth is also used to determine the strengthening methods and reinforcing depth of stone carving and grottoes. To obtain the weathered depth of Yungang Grottoes, an intelligent and efficient regression forecasting support vector machines (SVM) model based on terahertz spectra is established. Terahertz time domain spectroscopy (THz-TDS) is used to measure the terahertz spectra of the Yungang Grottoes. The transmission spectra of all samples are then analyzed. Only minor differences among the transmission spectra of different samples and no characteristic absorption peaks have been observed. The SVM model based on the transmission spectra of samples is applied to distinguish and predict the depth. The results are in accordance with the real depth, and the relative errors are less than 0.85%. The results indicate that the spectra-based model can distinguish samples of Yungang Grottoes and effectively predict their depths, which can provide an important reference for the repair of the Yungang Grottoes. Highlights: The rock core samples of Yungang Grottoes were measured by terahertz time domain spectroscopy. This method can facilitate the micro damage detection. The SVM prediction model was established based on THz transmission spectra of the samples.