【作者】 孟田华；

【导师】 杨进；

【作者基本信息】 中国地质大学（北京） ， 地球探测与信息技术， 2014， 博士

【摘要】 历经1500余年，作为世界文化遗产的云冈石窟，已经产生了严重的风化破坏。支持向量机（SVM）已在各个领域得到了广泛应用，但在石质文物保护领域中的应用研究尚未见报道。本研究首先利用太赫兹时域光谱（THz－TDS）技术获取云冈石窟石雕风化物在太赫兹波段的光谱数据，接着在云冈石窟实地进行超声波测试，进而通过波速比得到云冈石窟窟外壁的风化等级值。然后将两组数据结合，作为预测模型的输入特征向量，选择径向基函数作为预测模型的核函数，并利用留一法交叉验证（LOO－CV）寻得最优参数，建立一套基于光谱数据和SVM的石质文物风化等级预测模型。最后使用预测样本验证模型的可靠性以便推广使用。主要工作概况如下：1、利用THz-TDS技术研究了云冈石窟石雕风化物的THz光谱特性。实验结果表明，依据它们的THz透射或折射率光谱能够区分鉴别出不同石窟风化物。由此，可以提取出区分度较高波段对应的光谱数值作为SVM预测模型建立的训练和检测样本中的输入量。2、利用超声波（US）检测技术研究了云冈石窟各窟外壁测区的超声波波速，进而获得对应的风化等级值。并将各窟风化等级值作为SVM预测模型建立的训练和检测样本中的输出量。3、利用云冈石窟石雕风化物的THz光谱和US实验数据，以SVM基本原理和算法为基础建立的回归预测模型，可以从理论角度对石质文物的风化和损坏程度做分析预测。从预测结果可以得出，使用SVM方法对光谱数据具有较强的处理能力，即便光谱信息中没有特征峰，只要样品间光谱有微小差别，也可以实现鉴别并准确预测。4、使用X射线（X-Ray）衍射仪和扫描电子显微镜（SEM）对云冈石窟石雕风化物的进一步研究，不仅是对前面THz光谱分析和US检测的有效补充和辅助解析而且是对风化差异风化原因解释的有效佐证。5、在对国内外常见石质文物修复方法的总结下，提出针对云冈石窟适用的修护方法：将可控制光照进石窟进而实现控温的光学材料与能隔绝雨雪的保护性窟檐相结合的物理结构。更多还原

【Abstract】 With over1500history, Yungang Grottoes were recognized as a world heritagesite. It has suffered the serious weather denudationg from the natural and man-madeharmful effects. Although Support vector machine (SVM) has been widely used invarious fields, the applied research about stone relics protection has not been reported.Make use of terahertz time-domain spectroscopy (THz-TDS) technology, we canattain the THz spectral data of Yungang Grottoes weathered stone, field ultrasoundtests in this region, and then the weathered degree of outer walls of Yungang Grottoeswere obtained by wave velocity ratio. Afterwards we integrated two groups of data asthe input feature vectors of forecasting model, and chose radial basis function as thekernel function of prediction model, and selected the optimal parameters used theLeave-One-Out Cross Validation（LOO-CV）method, finally established a model forpredicting weathered degree of stone relics based on spectral data and the SVM. Theaccuracy rates by using the prediction sample and the Leave-One-Out CrossValidation before promote the use of model. All the works in the dissertation aresummarized as follows:1. By using THz-TDS technology, the THz spectral characteristics of YungangGrottoes weathered stone have been studied. The experimental results showed thatdifferent grottoes weathered stone can be separated or identified on the basis of theirTHz transmission or refractive index spectroscopy. As a result, the spectral valuescorresponding to wave band with higher distinguish degrees can be extracted as theinput of training and testing samples to establish SVM prediction model.2. By using ultrasonic (US) detection technology, the ultrasonic wave velocity ofthe outer wall test area in Yungang Grottoes have been studied and the correspondingweathering level value obtained. And the weathering level values of all grottoes arechosen as the output of training and testing samples date to establish SVM predictionmodel.3. Using the THz spectroscopy and US experimental data of Yungang Grottoesweathered stone, the regression prediction model could be established according tothe basic principle and algorithm of SVM. We can analyze and predict the weatheringand damage degree of stone relics from the theoretical aspect. The results turn out thatthe SVM method has a strong capacity for processing spectral data. It can alsoidentify and predict weathering level as long as there are small differences in the samples, even if there are no characteristic peaks in the spectrum information.4. The further study of The Yungang Grottoes weathered stone using X-raydiffraction machine and scanning electron microscope (SEM) is not only apromising complement and auxiliary analyzing for THz spectroscopy and USdetection, but also an effective evidence to explain the reason of the weatheringdifference.5. Based on the summary of both domestic and foreign stone relics repairmethods, the repaired method of combine the optical material with eave for YungangGrottoes was proposed. The optical material can control light and then realize controlgrottoes’ temperature, and the protective eave can isolate rain and snow.更多还原