Ethnic mountain settlements are living heritage of varied vernacular cultures.The preservation of both the built form and the intangible socio-cultural associations with them are global concern in process of urbanizat...Ethnic mountain settlements are living heritage of varied vernacular cultures.The preservation of both the built form and the intangible socio-cultural associations with them are global concern in process of urbanization,and in the notion of sustainable development.However,there is a lack of multi-dimensional and cross-cultural quantitative research in settlement morphology,making it difficult to guide practice effectively.Therefore,this study focuses on exploring an automatic or semi-automatic quantification and classification method for the morphological identity of ethnic mountain settlements.We introduce and combine 3-D morphological indicators with existing 2-D indicators to build and test three different sets of indication systems for semi-automatic classification for the settlements’ethnic attribute basing on spatial morphology.Taking the Miao,Dong,and Tunpu(Han)ethnic settlements in Guizhou province,southwest China as research samples,we applied factor analysis and hierarchical clustering methods to compare the classification accuracy under the three systems using data from topographic map,field investigation map,satellite imageries,and ethnography or local chronicle.The results showed that,the 3-D indication system has succeeded in semi-automatic quantification and classification of settlement morphology and ethnic identity by greatly increasing the classification accuracy to 96.30%,which is a huge improvement compared with the basic 2-D indication system(42.59%)and the advanced 2-D indication system(61.11%).The settlement samples are further divided into two sub-types with significant morphological differences in each major ethnic category under the 3-D indication system.We then discussed the potential improvement and future large-scale application of this method with the help of machine learning and other smart techniques.We hope to provide a comprehensive quantitative perspective and a more scientific reference for the future preservation and sustainable development of the massive and diverse vernacular heritages across the world.展开更多
基金supported by the National Key R&D Program of China(No.2018YFD1100303)。
文摘Ethnic mountain settlements are living heritage of varied vernacular cultures.The preservation of both the built form and the intangible socio-cultural associations with them are global concern in process of urbanization,and in the notion of sustainable development.However,there is a lack of multi-dimensional and cross-cultural quantitative research in settlement morphology,making it difficult to guide practice effectively.Therefore,this study focuses on exploring an automatic or semi-automatic quantification and classification method for the morphological identity of ethnic mountain settlements.We introduce and combine 3-D morphological indicators with existing 2-D indicators to build and test three different sets of indication systems for semi-automatic classification for the settlements’ethnic attribute basing on spatial morphology.Taking the Miao,Dong,and Tunpu(Han)ethnic settlements in Guizhou province,southwest China as research samples,we applied factor analysis and hierarchical clustering methods to compare the classification accuracy under the three systems using data from topographic map,field investigation map,satellite imageries,and ethnography or local chronicle.The results showed that,the 3-D indication system has succeeded in semi-automatic quantification and classification of settlement morphology and ethnic identity by greatly increasing the classification accuracy to 96.30%,which is a huge improvement compared with the basic 2-D indication system(42.59%)and the advanced 2-D indication system(61.11%).The settlement samples are further divided into two sub-types with significant morphological differences in each major ethnic category under the 3-D indication system.We then discussed the potential improvement and future large-scale application of this method with the help of machine learning and other smart techniques.We hope to provide a comprehensive quantitative perspective and a more scientific reference for the future preservation and sustainable development of the massive and diverse vernacular heritages across the world.
