基于大数据特征提取的建筑形态聚类检索方法研究——以大学校园为例

Research on Clustering and Retrieval Method of Building Form Based on Feature Extraction of Big Data: Take university Campus as an Example

Authors: Baizhou Zhang, Biao Li
Source: CAAD 2021, Wuhan, China
Date: September 26, 2021

PDF GitHub

Citation

Zhang, B., & Li, B. (2022). Research on Clustering and Retrieval Method of Building Form Based on Feature Extraction of Big Data: Take university Campus as an Example. Proceedings of the CAAD 2021, Wuhan, 25-26 September 2021, 606–613.

张柏洲, 李飚. 基于大数据特征提取的建筑形态聚类检索方法研究——以大学校园为例[C]//智筑未来: 2021全国建筑院系建筑数字技术教学与研究学术研讨会论文集. 2021: 606–613.

摘要

近年来,大数据技术在建筑及城市设计领域的相关探索与应用在不断更新发展中。其中,建筑形态相关的可视化数据如何抽象表征,并运用于设计分析与决策过程,是数据驱动的设计方法中的重要命题之一。本研究以大学校园作为研究对象,通过对地图大数据的筛选与要素提取,收集建筑形态布局相关的量化特征,借助计算几何、神经网络等技术手段,研究建筑空间形态角度的案例匹配与支持设计决策的可行性。研究首先从网络开源地图平台对大学校园数据样本进行采集和筛选;进而通过几何规则算法、图像特征提取等方法从地块形态、建筑形态、道路形态三方面分别对大学校园样本进行形态特征的提取;最后将形态特征整合,根据综合特征数据和聚类结果进行近似案例匹配的实验验证。研究结果表明,该研究方法能够将建筑形态要素进行合理的量化表征,并从形态角度将建筑样本进行分类并探究一般规律,从而进一步助力设计者从大量数据中快速获得较为可靠的近似案例检索以支持设计决策。

关键词

特征提取;大数据;建筑形态;大学校园;聚类;案例检索

Abstract

In recent years, the relevant exploration and application of big data technology in the field of architecture and urban design have been continuously updated and developed. Among them, how to abstract represent the visual data related to architectural form and apply it in the process of design analysis and decision-making is one of the important propositions in data-driven design methods. This study takes university campus as the research object, through the screening of map big data and element selection, collects the quantitative features related to the architectural layout, and studies the feasibility of case matching and supporting design decisions from the perspective of architectural morphology with the help of computational geometry, neural network and other technical means. Firstly, the data were collected and screened data open source map. Secondly, through the computer geometry and neural network technology, the feature extraction methods were performed from three aspects: site, road and building. Finally, different aspect of features was integrated and an experiment of matching approximate cases according to the features and clustering results was conducted to validate the research methods. The results show that the research method can characterize the architectural morphology elements reasonably and quantitatively and classify architectural samples from the perspective of morphology to explore the general law, thus further helping designers to quickly obtain more reliable approximate case retrieval from big data to support design decisions.

Keywords

feature extraction; big data; building morphology; university campus; clustering; case retrieval