Research On Measurement Method Of Occupants’ Spatial Behavior For Post-occupancy Evaluation Based On Computer Vision Technology

Post-occupancy evaluation(POE)is a user-centered evaluation method,which focuses on how users use the built environment.It is one of the important ways for architecture to obtain post-occupancy performance feedback.At present,the main methods for collecting user activity data in POE,comprising mainly observation,interviews,and questionnaires,are subjective,uneconomical,inefficient,small in sample size,and difficult to quantify.Therefore,this paper develops a measurement research method that introduces sensing technology,computer algorithm and machine learning to quantitatively describe the spatial behavior of users in the built environment and provide quantitative measurement technology for POE.On the premise of not interfering with or being interfered by users,this method can automatically and continuously record and describe the spatial behavior of users in the space in the whole life cycle of the built environment,and can obtain the high-resolution spatial behavior big data and quantitative indicators of anonymous behavior variables based on time more accurately and efficiently than similar methods.In terms of the theoretical support,this paper comprehensively reviews and summarizes the literature related to users’ behavior,and puts forward a concept and definition of "occupants’ spatial behavior" oriented to architecture,which defines the scope and establishes the theoretical basis for the research.33 most valuable POE evaluation frameworks around the world are reviewed,and their evaluation objectives,evaluation tools and data collection methods are arranged,classified and analyzed,and the limitations of these protocols are deeply analyzed in practical application from the perspective of methodology.This paper constructs a comprehensive measurement and evaluation framework of 7 variables and 82 quantitative indicators to quantify occupants’ spatial behavior.At the same time,various modes of occupancy,movement and status are introduced in detail.At the level of technical tools,a measurement method of quantifying occupants’ spatial behavior based on computer vision technology is developed.The method is divided into three major steps: first,we collect occupants’ spatial behavior image data to train and obtain a reusable computer vision model of the built space;Secondly,user image data of the space to be evaluated is processed by the computer vision model,undergoing affine transformation to obtain a standardized,large data set of spatial behavior locations;and third,the standardized data set is algorithmically processed to obtain quantitative POE indicators of spatial behavior.At the practical application level,the reliability of this method in practical application is verified.The accuracy,applicability and convenience of the developed method are verified in a series of real space behavior POE case studies,such as the safety evaluation of kindergarten children’s outdoor activity behavior,the evaluation of the use efficiency and occupancy behavior of indoor atrium space in university library,and the evaluation of the support of outdoor square facilities on university campus for resident behavior.This method can adapt to various spaces,has the advantages of simple operation,low cost,objectivity,economy,quantification and generalization,and has potential competitiveness for large-scale popularization and application.It introduces computer vision to the POE domain,provides it with an economic and reliable quantitative measurement method,and also provides a datadriven basis for improving spatial sustainability,planning,design,and decision making.