IIC Journal of Innovation 15th Edition | Page 20

Physical Distancing and Crowd Density Monitoring
peak violation during peak café times that is between 11:45 am to 12:30 pm . Surprisingly , peak violation (%) is during 1 pm to 1:30 pm time slot .
OPERATIONALIZATION
Typical operationalization involves setting up the real-time component in an on-premise environment and the back-end component in cloud :
On-premise component : This includes the SAS ® Event Stream Processing ® engine configured with a deep learning model and real-time dashboard . Setup includes configuring connectivity to Video cameras generally via a Video Management System ( VMS ) for inbound data . The target environment is typically an Edge server with GPUs to process the video analytics workload . It is possible for a single edge server to support multiple cameras depending upon the compute available . Additional edge servers can be added to handle a high number of cameras . Back-end component : This is the SAS Visual Analytics ® application deployed in a data center or cloud environment . Since the videos are processed on-premise , the payload and storage requirements are modest and can be easily scaled .
Fig . 8 : An interactive dashboard for time and location-based analysis . The bar char on the top left provides the distribution of customer arrival and violation (%) based upon the day of the week . It is observed that more customers arrive on middle of the week compared to start or end of the week . Violations follows the same trend . The tree chart on the top right gives insight about number of individuals visiting a give geofence and the observed violations in the location by the size of the section and the color of section respectively . While more individuals have visited the cashier section compared to the dessert section , the violation percentages are higher
IIC Journal of Innovation - 15 -