Font Size: 

Introduction

The goal of this research is the description of spatiotemporal of occupancy patterns in specific building settings: chirurgical hospital corridors. Until now, post-occupancy studies in architecture have focused on occupancy, movements and interactions, demonstrating correlations of spatial descriptions with global patterns of human behavior. However, no studies had focused on the influence of space in relation to the activity nature to the occupancy patterns. The hypothesis is that patterns of occupancy will emerge from the programming of activities due to the influence of space. The challenge is to capture occupancy data with high temporal and spatial resolution, accuracy and precision that allows us to ask relevant research questions regarding time.

 

The exploration of occupancy patterns depends on the characteristics of the data collected. If data is low temporal resolution, or if temporal variable is merged to become constant, there could be no conclusions about the influence of the programming of activities on the outcome. Therefore, this paper presents the use of vision-based techniques to explore spatial and temporal occupancy patterns, using an empirical study as proof of concept, conducted in a specific spatial setting: An academic corridor. This paper describe the method to obtaining an accurate and high-resolution dataset that allows us to depict the technical, methodological, and social implications of analyzing data captured in specific spatiotemporal scenarios; the spatiotemporal data; and the exploratory analysis proposed.

 

Background

One of the most important models for spatial descriptions, developed in the 80’s by Hillier and Hanson (1984), is Space Syntax theory, which creates diverse representations of the geometrical components of the space units and their relationships. One of the major contributions of this area is not only the description of space through more abstract attributes, but also their correlation to human social behavior (Peponis, Zimring, Choi, 1990). From this perspective, researchers have focused on effects of spatial layout on human behavior, with most emphasis on occupancy, movement, and interactions. To collect the necessary data of human behavior, they have utilized methods that range from self-report to direct observation and mapping, both requiring human interpretation (Weisman, 1981; Moeser, 1988), and the last one being more expensive in terms of resources, requiring the continuous presence of an observant. These methods have been utilized to respond to research questions that are limited to spatial dimensions, disregarding the temporal one.

 

Physical location and tracking has been the focus of computer science research for a several years. They have developed systems that range from sensors and cameras to hybrids systems, such as Kinect sensors. Several of them can be applied to Architecture research. Vis-A-Viz tool, developed by Romero (2008), for example, is a tool for visualizing activity through computer vision. It captured data from overhead videos of a research home-lab environment. It automatically records and processes human movements from overhead videos captures, allowing a close one-to-one mapping of individuals’ position over the architectural layout. A computer vision approach to capture aggregated movements by hour, in ‘informal learning spaces’ (Tome and Heitor, 2012), presented a correlation between movements and ‘space-use’ analysis, specifically axial lines (Penn, 2009), using ArquiTracking application. Gomez, Romero, and Do (2012), used three overhead high-resolution video cameras to collect data for the study of Activity Shapes, capturing three activity scenarios. This study demonstrated that the nature of the activity does influence the distribution of occupancy on a space.

 

Methodology

For our study, we collected finer grain spatial and temporal occupancy data to analyze a specific circulation scenario. We used security cameras to collect high-resolution data testing several settings. The case presented in this paper corresponds to the data captured with a resolution of 30fps, during 1 week. We developed semi-automatic computer-vision methods that allow us extract the individual’s coordinates in space and time (x,y,t). This is the central part of the research, since it provides a methodological platform for a long-term research program, which final goal is to explain patterns of occupancy and the impact spatiotemporal factors on activities, such as to what extent the geometry of the setting matter, separating out answering the later conceptual questions regarding spatiotemporal patterns from the methodology that allows answering them. The methodology from the data collection though data processing and exploratory outcomes are the fundamental part of the expected outcomes of this research.

 

Contribution

The expected contributions of this research are: (1) Methodological contributions about the procedures for capturing high-resolution spatiotemporal data, which is the initial step of a long-term behavioral research program. (2) Methodological contributions about the procedures for data processing and exploratory analysis of the data obtained. (3) And, outside the architecture field, a high level taxonomy of activities is expected to contribute to the area of Activity Recognition of computer science, since their research is mostly based on decomposing activities, but does not present a higher-level framework.

Computer Vision; Data collection; Occupancy patterns; Spatiotemporal Analysis; Patterns description