Summer Research Internship Project 2018 – Sun Park


Sun Park is a third-year student at the Faculty of Liberal Arts and Professional Studies at York University. Having specialized in Geography and Urban Studies, Sun is spending the summer in the GeoICT lab, under the supervision of Dr. Gunho Sohn, exploring 3D modelling and pedestrian comfort analytics in a highly active area on campus. Specifically, Sun will be utilizing a virtual 3D model of the student centre she created to visualize tracked pedestrian movements in a campus setting. By the end of the summer, Sun is hoping to have furthered the research to gain a better understanding of how pedestrian comfort is influenced by highly active environments and spaces with regards to urban design. This research is important because it considers an individual’s level of comfort in areas of high activity and is valuable for producing good urban design and spaces.


Pedestrian Comfort Analytics Using Surveillance Camera and Dynamic Virtual Campus Models

Understanding pedestrian comfort is a crucial aspect for urban designers and planners to produce what is known as “good” urban spaces or environments, especially in areas of high pedestrian activity. Moreover, the quality of life tends to rise in these areas when urban design and city planning considers an individual’s level of comfort. Currently, knowledge on pedestrian comfort with respect to urban design at York University is limited, specifically in terms of the student centre colonnade. This pedestrian walkway is one of the commonly used spaces on campus that displays various activities. The objective of this research is to develop a virtual world for tracking pedestrians from surveillance video sequences installed in the colonnade, and generating animated characters to represent tracked pedestrians for gait visualization. Therefore, it is primarily focused on analyzing pedestrian movements in order to understand a degree of comfort in a highly active area. A 3D model of the student centre was created using LiDAR point cloud data as a reference. The usage of pre-trained deep learning algorithms will allow for detecting and tracking pedestrians. The detected pedestrians are transferred from 2D images into 3D virtual animated characters, including footsteps and body poses in 3D. This allows us to analyze the spatio-temporal parameters of poses for comfort measure for each individual. We will present analytics results obtained by several visualizations based on the results from deep learning based tracking. This will give us an opportunity to use computer vision and machine learning as a tool for gaining knowledge on pedestrian comfort with regards to urban design.