Video analytics process camera images to send alerts out so operators can respond to specific incidents or collect data for historical trend analysis. We have seen an increase in the types of analytics offered over the past several years and the power of the hardware that supports it.
Analytics will always be supplemental as they require the video management system (VMS) and potentially an access control system (ACS) as the backbone for their processing and referenced libraries.
They are an additional layer that resides on top of each camera or the VMS architecture. Video analytics function by either comparing the series of images to a library for recognition or by measuring and detecting the change of pixels to determine direction or locations. This is the foundation of technologies such as:
- Image recognition
- Object identification
- Object tracking
- Motion detection
- Directional trip lines
- Dwelling zones
- Slip and fall detection
- People counting
This is not an exhaustive list but knowing what is out there is the first step to knowing what can be supplemented in a security program.
Edge or Central Processing
Two video analytic structures are edge processing and central processing in the Cloud, or through an on-premises server. When looking at analytics, infrastructure bandwidth and data storage are critical to consider as each type will have different requirements.
The edge-based analytics process images at the camera and can send the results of the analysis to central repositories, video management systems or alerting systems to generate an alarm or store the data for historical use. While this does lighten the bandwidth required as less information is crossing the network, the camera may be storing data locally.
If a security program has hardened cameras, this may not be a substantial risk. However, if the cameras are left relatively open, a bad actor could access the stored analytical data or find out where the data is headed.
Patching and updates can be challenging as the camera firmware version, the VMS version and the analytic version all need to be verified for compatibility. If a new iteration or patch comes out, the end user or their integrator will need to verify all three are compatible before updating. This is a great solution if bandwidth is a concern and the cameras you have support this level of processing power.
Central processing takes incoming video footage, processes it and integrates with the VMS to send out alarms. This requires more bandwidth across the network as these will either require an additional stream from the cameras or can act as a midpoint from the cameras through to the VMS. The architecture and hardware costs can make scaling in an enterprise environment challenging if an on-premises server can process 20-30 cameras each.
Some analytics servers allow the user to add video streams and push updates from a central location much like a VMS. If there are any patches required for the analytics, the user will need to update the servers instead of reaching out to each individual camera.
Deployment and maintenance costs are a substantial factor in determining a solution’s viability within a security program. The initial deployment of the technology involves identifying which cameras will receive the analytics, the configuration standards and SOP documentation around the generated alarms.
The more complex the analytics, the more man-hours it will take to configure or calibrate for optimal use. As the technology piece is finalized, operations teams must write and conduct the training on SOPs for operators. This may be quickly amending current alarm-based SOPs or writing new SOPs for a newly defined incident.
Post-deployment, periodic maintenance is an important operational expenditure to account for. A system upgrade or firmware vulnerability patch could require an administrator to touch each camera. Depending on the size of the system, updating each camera and the analytics could turn into a biddable project. This labor is important to consider when looking at deploying a video analytic solution, however, it is equally important to determine the cameras that will get it.
Targeting the analytics that specific cameras will receive and identifying whether your infrastructure or cameras can support the analytics’ architecture will increase your program’s security posture while driving a positive ROI. Video analytics are typically not for every camera in your system but can be a powerful instrument in today’s security technology toolbelt.
Josh Akre is Performance Engineering Manager at Northland Controls.
The post Understanding Video Analytics: Edge vs. Central Processing Structures appeared first on Security Sales & Integration.