IIC Journal of Innovation 15th Edition | Page 14

Physical Distancing and Crowd Density Monitoring
the entry way ( lower-left ). We also send out real-time alerts when capacity limits have been exceeded using ESP which is discussed in the next section .
Event Stream Processing
SAS ® Event Stream Processing ® ( ESP ) enables real-time execution to analyze many continuously flowing events . ESP can be deployed on edge servers , on-premise servers or in the cloud . In this project , ESP was used for the following :
- Read the streaming images from the IP camera - Pre-process the input image streams - Detect person using the pre-trained computer vision model - Track the detected people across frames - Perform post-processing to calculate distances between the detected persons - Define geofences and track people in them - Send alert notifications via email or SMS to the admin in case of any violation - Feed the data for both real-time streaming dashboard and post-facto dashboard
Real Time Analytics
Figure 4 displays the contents of the real time streaming dashboard which is populated by the data coming from the ESP engine . The dashboard has various components such as streaming video from the ESP server with markers indicating the distance between two individuals . The image viewer also contains the transformed view which is the image after applying perspective transform . With streaming running in real time , we also have live analytics such as the total number of individuals in each geofence at every point of time . The alert table in the dashboard has the list of violation alerts generated by ESP server .
IIC Journal of Innovation - 9 -