Research On Methods Of Clustering And Quantification Of Chinese University Campus Morphology

University campus,as a product of urban development,possesses a long history of development and evolution.With the worldwide wave of higher education development in the 1960 s and the peak of university planning and construction in China since the new century,an increasing number of domestic and foreign scholars have been conducting research on university campuses and their forms.Nevertheless,most existing studies on the morphology of university campuses are based on morphological evolution and typological patterns,with a preference for qualitative generalization and subjective descriptions of university campus morphology and its prototypes,but with a lack of quantitative research and explanatory analysis supported by data.Accordingly,this project refers to the quantitative analysis methods emerging from urban morphology and integrates data crawling and data processing techniques in the context of big data,neural network models in the field of deep learning,image processing and clustering algorithms in the field of computing,and proposes an extensible research framework for the analysis and quantitative study of university campus morphology: a Chinese university campus figure-ground dataset is constructed based on multi-source data(Including information data,POI data,map data and customized figure-ground data of campus),and two clustering experiments based on the dataset are conducted to explore the feature representations of university campuses and to quantify the similarities and differences of different categories of university campus morphologies.The purpose of this paper is to provide data support and theoretical guidance for future campus morphology research and campus design,as well as to quantify the characteristics and morphology of more functional types of buildings through an emerging quantitative framework in this paper.Chapter 1 of the thesis is an introduction,which illustrates the background,purpose and significance of the study.Chapter 2 explains the concept,development history and theoretical research of university campuses and their morphology,with a technical introduction of the neural network modelling algorithm model,while pointing out the great potential of quantitative methods of urban morphology in the quantitative study of campus morphology.The third chapter systematically describes the composition,construction method,parameter description and application value of the Chinese university campus floor plan dataset from the perspective of multi-source data acquisition.Chapter 4 and Chapter 5 focus on "shallow index feature quantification" and "deep image feature quantification" respectively,and try to cluster university campuses through two clustering experiments with different research frameworks and experimental methods.In Chapter 4,six types of feature factors are measured on the basis of the dataset to describe the morphology of college campuses,and the statistical description,correlation analysis and clustering of the six types of feature factors are conducted through statistical analysis;while in Chapter 5,the deeper features of the customized floor plan of college campuses in the dataset are extracted through the neural network algorithm model of convolutional self-encoder,so as to conduct the clustering study and quantification of differences.Drawing on the previous section,Chapter 6 compares the research framework and findings of the two clustering experiments,and discusses their differences,advancements and limitations with the traditional campus morphology classification approach and the quantitative method of urban morphology.Full text of about 65,000 words,94 diagrams(including 9 quotes,8 repainted,77 self-painted)...