University Researchers Evaluated Pedestrian Distancing Behavior Using Traffic Cameras Near Intersections in New York City.
New York City, New York, United States
Summary Information
This study’s primary objective was to introduce a continuous, remote, and real-time pedestrian detection and social-distancing data acquisition system on an existing video surveillance system in New York City. It provided a deep learning-based methodology to detect pedestrians, vehicles and cyclists with a generalized real-pixel distance rate method to the approximate real distance between a pair of pedestrians regardless of the different perspective views of the cameras. The project examined temporal and spatial changes in pedestrian density and social distancing behavior at multiple locations in the five boroughs of NYC using 68 traffic cameras.
The estimated deployment costs were summarized for key components in deploying the pedestrian detection system.. Researchers assumed that video feeds would be continually collected in 30-second intervals. Costs based on local server and cloud-based services were both estimated. For cloud-based services, annual costs were based on a 3-year-long system deployment and the plain file storage method (where images were stored directly to the server). Cloud-based costs include cloud server service with 2 CPUs and 1 gigabyte (GB) of memory, data storage and data transfer service.
Local Server-based Services |
Cost component |
Cost |
One-time server acquisition |
$600 |
|
Network Attached Storage (NAS) device (up to 65TB) acquisition |
$4390 |
|
Annual Cost (Based on a 3-year estimate) |
$1663 / year |
Cloud-based Services – Option 1 |
Cost component |
Cost |
Cloud server (2CPU, 1GB memory) |
$67.56/year |
|
Data storage with infrequent access |
$394.2/year |
|
Data Transfer (180 GB/per month (Inbound) and 50 GB/per month (outbound)) |
$52.92/year |
|
Annual Cost (Based on a 3-year estimate) |
$514.68/year |
Cloud-based Services – Option 2 |
Cost component |
Cost |
Cloud server (2CPU, 1GB memory) |
$67.56/year |
|
Standard data storage |
$906.72/year |
|
Data Transfer (180 GB/per month (Inbound) and 50 GB/per month (outbound)) |
$52.92/year |
|
Annual Cost (Based on a 3-year estimate) |
$1,027.20/year |
The costs were projected to vary depending on the number of cameras supported and data collection interval timing. The local server-based service was projected to be cheaper in the long term, provided the data storage did not exceed the data storage device limit (e.g., $998/year for a 5-year estimate). Cloud-based services were more expensive in the long term as more data needed to be stored on the cloud (e.g., $1,631/year for a 5-year estimate if using S3 standard storage).
Reference-Free Video-to-Real Distance Approximation-Based Pedestrian Detection System Amid COVID-19 Pandemic
Local Server Services (68 Cameras): $1,663 per year for three years.
Cloud-Based Services (68 cameras): $1,027 per year for three years.