(iPoster) Quantifying Global Surveillance Camera Distributions from Street View Images

• In Event: Poster Session: Information Technology and Politics & Science, Technology, and Environment
  In Poster Session: Poster Session: Information Technology and Politics

Abstract

Public space surveillance cameras epitomize digital surveillance, yet their global distribution remains poorly understood.
We present the first systematic, scalable, and transparent estimation of public space surveillance camera density.
We utilized nearly 18 million street view images randomly sampled at 2.95 million GPS coordinates from the 1,630 most populous cities in 178 countries from 2007 to 2022. A deep learning algorithm was developed to detect the presence of surveillance cameras in these images.
We measured surveillance camera density as the proportion of locations in a region where the algorithm could identify cameras.
Our analysis reveals three key findings. First, East and Southeast Asia lead in both current camera density and installation rates, while Western Europe and North America---the most studied regions---lag behind.
Second, countries with higher GDP per capita have more cameras currently but show slower growth in camera installation rates; regime type does not predict camera prevalence but impacts where cameras are installed in cities.
Third, leveraging our dataset's panel structure, we find that increases in camera density within countries predict more robberies and protests, contrary to cross-sectional studies suggesting cameras reduce these activities.
These findings challenge conventional wisdom about surveillance cameras' effectiveness and distribution. Our publicly available dataset, CADESI, provides a foundation for empirical research on digital surveillance's causes and consequences.

Author

• Han Zhang, Brown University