Physical Distancing and Crowd Density Monitoring
close ( yellow or red , respectively ) versus people that are isolated ( green dots ). This results in a graph where we can consider the connected components of the graphs to be clusters that need to be monitored and tracked over time . We can also keep track of the number of clusters and the size of each cluster over time . As part of our summary analysis we can focus our attention on clusters of a given size that persist for longer periods of time between the same people as shown in Figure 3 below .
Fig . 3 : Real-time visualization of person-detection , tracking , and geofence counting , and clustering . On the left , we have the original camera footage with gray lines depicting userdefined geofences . In this example , we have four geofences corresponding to the food stations ( top-right ), the dessert area ( lower-right region ), the cashiers ( rectangular region in the middle ), and the entry way ( lower-left ). A dot is placed on each detected person . After applying a perspective transform for camera calibration , we show on the right the location of people and the estimated distances between two people when they are closer than 6 feet apart and draw an edge between them . The edge is colored red if their distance is less than 4 feet , and colored yellow if the distance is less than 6 feet . On the bottom right , we summarize information about the clusters and the density of people in each geofence .
Density Analysis with Geofences
We also perform a density analysis with user-defined geofences . The user defines regions of interest along with capacity limits ( optionally ) for each region . Our solution counts the number of people detected in each region at every time-point . On the left half of Figure 3 , we show our implementation in the example of our SAS café camera . The gray curves show the user-defined geofences . In this example , we have four geofences corresponding to the food stations ( topright ), the dessert area ( lower-right region ), the cashiers ( rectangular region in the middle ), and
- 8 - November 2020