基于机器学习的历史空间感知测度研究  

Applying Machine Learning in Measuring the Perception of Historical Space

作　　者：曹越皓 杨培峰 庄凯月 Cao Yuehao;Yang Peifeng;Zhuang Kaiyue

机构地区：[1]重庆大学建筑城规学院 [2]福建工程学院建筑与城乡规划学院

出　　处：《规划师》2021年第23期67-73,共7页Planners

基　　金：国家重点研发计划项目(2019YFD1100704);国家自然科学基金项目(51878084)。

摘　　要：历史空间感知是人与历史环境要素相互作用和影响后形成的整体感受与认知,是从人本视角研究历史遗产的重要方向。虽然在当前的历史遗产保护研究与实践中已形成较为完善的物质空间保护框架,但是针对人与历史空间交互关系的研究较少。而传统研究方法受成本高、样本量少、偏差较大的限制,很难对大规模或大尺度的历史空间感知进行高精度测度。基于此,文章融合历史空间互联网评价的多源数据,利用时空数据计算、自然语言处理和计算机视觉等机器学习技术方法构建历史空间感知测度技术框架,从空间结构分布、空间类型聚类、典型意见抽取、情感倾向判断、图像内容识别和历史意象归纳6个维度对历史空间感知进行测度研究,并以舟山为实证案例,分析其历史空间的感知特征与核心问题,以期为舟山历史文化名城保护规划的编制提供方法支撑。Historical space perception is formed through the interaction and influence of human and historical environment elements,and it is the most direct way to understand and experience urban history.Although a framework for historical material space has been formed in the conservation planning of historical and cultural cities,research on the interaction between people and space is rare.Meanwhile,traditional survey methods are limited by high cost,small sample size and large deviation,so it is difficult to measure large-scale historical spaces with high precision.Based on the Internet comments data of Zhoushan historical and cultural city,this paper constructs the perception model of historical space by using GIS spatial computing,natural language processing(NLP)and computer vision(CV),and measures the perception of historical space from six dimensions:spatial structure distribution,spatial type clustering,typical opinion extraction,emotional tendency judgment,image content recognition and historical image induction.The result summarizes the characteristics and problems of Zhoushan historical spaces,and provides an important support for the conservation planning of Zhoushan historical and cultural city.

关 键 词：机器学习 历史空间 感知测度 历史文化名城 舟山 

分 类 号：TU984.113[建筑科学—城市规划与设计]