摘要
公共空间是当代城市的重要组成部分,基于人群空间行为的研究可为其优化设计提供参考。当前,利用视频数据获取行人轨迹进而评估公共空间的研究方法开始出现,但使用的方法具有运算速度慢,无法实时获取结果等缺点。使用Python编写基于深度学习的计算机视觉算法,可实时获得研究场地上行人的轨迹数据。以天津大学卫津路校区内的3个公共空间为例,利用轨迹数据绘制人群分布热力图和人群行走速度热力图来分别表征公共空间各个出入口间的连接强度,以及行走速度不同的行人所选择路径的空间分布差异。最后,提出了该方法在辅助设计方面的可能性及其存在的局限性和改进的策略。
Public space is an important part of contemporary cities,and researches based on human spatial behavior can provide a reference for its optimal design.Currently,research methods which obtain pedestrian trajectories with video data to further evaluate public spaces are emerging.However,the current methods have the disadvantage of slow computing speed and the inability to obtain results in real time.A computer vision algorithm based on deep learning using Python can obtain the trajectory of pedestrians in the public space in real time.Citing the three public spaces in the Weijin Road campus of Tianjin University as examples,this research uses the trajectory data to draw a heat map of crowd distribution and a heat map of walking speed to respectively characterize the connection strength between the entrances and exits of the public spaces and the spatial distribution differences of the chosen paths of pedestrians with different speeds.Finally,it puts forward the possibility of this method in assisting design,its limitations and improvement strategies.
作者
吴韶集
胡一可
WU Shaoji;HU Yike(School of Architecture,Tianjin University;Architectural Design Research Center Studio 11, Tianjin University)
出处
《风景园林》
2022年第2期106-111,共6页
Landscape Architecture
基金
国家自然科学基金(编号52038007)。
关键词
深度学习
计算机视觉
城市公共空间
行人轨迹
无人机
deep learning
computer vision
public space of city
human trajectory
unmanned aerial vehicle(UAV)
作者简介
吴韶集,男,天津大学建筑学院在读硕士研究生,建筑设计研究中心第11工作室成员,研究方向为公共空间与人群行为;通信作者:胡一可,男,博士,天津大学建筑学院副教授、博士生导师,本刊特约编辑,研究方向为风景园林规划与设计、城乡公共空间与人群行为、风景旅游区规划设计,邮箱(Corresponding author Email):563537280@qq.com。
