BriefCam People Counting white paper
Introduction
One particularly important video analytic capability is people counting.
BriefCam offers two types of people counting:
Flow counting – The ability to count the number of people that pass through a scene or a predefined section or path in the scene.
Concurrent counting – The ability to detect the number of people in a pre-defined area or view.
People counting is especially helpful for gathering business metrics such as in retail scenarios, where it can be used to monitor queues in order to prevent checkout abandonment by mobilizing staff based on real-time indicators; at live events, stadiums, campuses or any venue where preventing crowding and optimizing traffic flows is inherently connected to customer satisfaction; and in waiting areas where preventing long waiting times could directly impact customer retention and revenue.
Concurrent counting can be used to configure rule-based alerts to trigger notifications when certain pre-determined threshold quantities of people have been exceeded. For example, a count-based alert can be used to trigger when there are five or more people in a checkout line, requiring another checkout to be staffed. An alert can also be used to trigger security when there are too many prisoners at one section of a prison or too many people at the park at night, indicating a potential riot.
Flow-based counting can be used, for example, to see how many people enter a store at different times (to be analyzed within RESEARCH) or alert when more than a predefined number of people have entered a hallway in an hour. In addition, an alert can be used to enable the timely flow of people, such as in an airport, where an alert can call a steward to assist, and can be used to maintain hygiene standards by triggering maintenance staff when a certain number of people have visited the restrooms.
Flow-based counting can also be used to approximate occupancy by counting all the people entering a space (by monitoring the entry ways) and how many exit a space (by monitoring the exits) and aggregating the results.
What You Can Count
In BriefCam, you can count:
People in a scene or an area within a scene.
People that spend a certain amount of time in the area.
People that cross a single or multi-segment line in a specified direction.
People that travel along a defined path.
People with specific attributes in addition to the above (using the whole array of BriefCam filters)
Limitations of People Counting
Note
The statistical accuracy for large count numbers is up to ±10%, but it may be larger for smaller counts or in challenging scenes.
Every location is different and, therefore, it is important to remember that calibration of the cameras is necessary for successful people counting. Therefore, it may take some time to find the ideal position to give the best results.
People counting relies on the object extraction and tracking mechanisms – which are not perfect and counting exacerbates these imperfections, especially when there are low counts of people. Therefore, 100% accuracy is usually not achievable unless the scene is very controlled and sterile.