Authors: Fan Zhang*, Massachusetts Institute of Technology, Di Zhu, Peking University, Yu Liu, Peking University
Topics: Geographic Information Science and Systems, Urban and Regional Planning, Environmental Perception
Keywords: Deep Learning, Urban Mobility, Street View Image, Social Sensing
Session Type: Paper
Start / End Time: 3:05 PM / 4:45 PM
Room: Washington 6, Marriott, Exhibition Level
Presentation File: No File Uploaded
Street-level imagery have been blanketing the all-around landscape of an urban area. This new source of data shows its advantage in fine-grained earth observation when compared with satellite imagery regarding not only physical environment survey but also social sensing. Prior works using street-level imagery have primarily focused on urban built environment audit. In this study, we demonstrate the potential of street-level imagery in uncovering urban mobility patterns. Our method assumes that the streetscape depicted in street-level imagery reflects urban functions, and urban mobilities in streets of similar functions have similar temporal patterns. We predicted the hourly variance of taxi trips of a street with a single street view image using a deep convolutional neural network and achieved outstanding performance (OA = 87.11%, MAPE = 32.98%, R^2 = 0.841). The study shows that street-level imagery, as a supplement of remote sensing imagery, provides us an opportunity to derive fine-scale socioeconomic information of an urban area and could therefore help urban planners and social scientists in urban observation, urban studies, and planning.