Quantitative Analysis For Deterioration Risk Of Ancient Grotto Murals

The new generation computing techniques have significantly boosted human’s capability of data perception,analysis and decision,thus have promoted the profound fusion of multiple related disciplines,from which digital humanities and computational social science have emerged recently.The conservation of cultural heritages is a typical new interdisciplinary area,which does not only rely on computational methods to improve the efficiency of domain work,but also,due to the specialty of difficulty,present huge challenges to computational techniques.Aiming at risk quantitative analysis,one of the most important problems of protective conservation of cultural heritages,this thesis focuses on ancient grotto murals and has done the following four aspects of investigations.First,lacking of suitable risk quantitative analysis theory is a fundamental problem for the protective conservation of unmovable culture heritages.Based on the classical risk management theory in other areas,this thesis concentrates on the unique deterioration risk of cultural heritages and proposes a complete risk quantitative analysis scheme for grotto murals.Second,reliable monitoring data of murals is the key to realize accurate risk loss evaluation and correlative factors analysis.This thesis uses laboratory aging-simulation and builds a multi-channel full-life-cycle image benchmark data of disruption,a typical deterioration of murals.This thesis also proposes an effective deep convolution network for mural images based deterioration grading.The practice of risk level assessment of grotto murals is presented.Third,for long-term yearly status census data of ancient murals,due to the huge data size,inconsistent quality,high dimensionality,multi-modality,heterogeneity,complex spatial-temporal property,unknown deterioration developing mechanism,it is impossible to realize automatic joint analysis for such kind of data with data mining.This thesis proposes a satisfactory visualization method to analyze the internal factors in spatial-temporal domain of deterioration risks.A set of joint visual analysis tools for deterioration risk and internal factors,such as spatial distribution,chronology,area,value,are developed,supporting multi-view,multi-scale and multi-facet analysis for domain experts.Fourth,for the multi-source complex monitoring data of murals,environmental,micro-environmental factors,domain experts cannot realize effective correlative analysis among mural deterioration risks and external factors.This thesis divides this challenging task into two sub-problems: 1)analyzing the influence of external environmental factors(temperature and relative humidity)to the hosting micro-environment of murals,and 2)analyzing the correlation of micro-environmental factors to mural deteriorations.Relying on visual analysis and clustering,this thesis presents a convenient joint analysis approach for the real-world monitoring data of mural deterioration developing speed and related environmental factors.This approach can support and validate some theoretical judgments about the mural deterioration development and related environmental factors,and from data quality and acquisition frequency,rationalize future monitoring plan of cultural heritage sites.The proposed risk quantization analysis scheme,deep CNN based mural deterioration grading approach,and joint visual analysis of mural deterioration risk and internal/external factors have already been applied in Dunhuang Mogao Grottos,which is the first valuable trial of using data-driven quantitative risk management theory and hybrid intelligence based on deep learning and visual analysis to solve real-world challenging protective conservation problems of cultural heritages.