Security cameras, now number in the tens of millions in use today, are proliferating life and how we surveil our businesses and even our homes. They are currently producing 500+ petabytes of data per day (Petabyte = 1,000 Terabytes).
Most of the security cameras installed are not watched by a human or reacted to in real time, they are for archival or investigative purposes after the fact. However, they are getting smarter in what they are able to 'automatically' render and provide alerts to. This is called video analytics and it's a simple math problem (relatively) comparing pixels from one second to the next. Most Video Management Systems (VMS) come with some basic video analytics. Object left behind, Object removed, wrong way, tailgating, virtual fence, etc.
First off, most security cameras are processing images just the same as the camera on your phone, but they have a way to save on data storage by taking a snapshot (called the I-Frame) and then only updating the pixels on the recording or screen that have changed in milliseconds (P-Frames). If you recorded/transmitted all I-frames it would be a larger amount of data and processing since you are refreshing all of the pixels in every image recorded. I Frames should typically be refreshed every 4-6 images depending on the content being observed. This reduction in full images saves on storage, processing power, and bandwidth required on the network and is commonly known as video compression. This compression standard H.264 is in 90%+ of the applications out there and can handle up to 8k resolutions. Fun Fact; this is also how you receive the majority of your streaming content and television (Jeopardy won't be the same).
Back to video analytics we can use simple programming math and comparison of the last I + P Frames to alert an operator or software of objects (bags, people staying too long near a restricted area) that are left behind for more than a specific amount of seconds, or if an object is removed let’s say in an art gallery you can immediately trigger an alarm. In an airport you can easily compare pixels that are moving in a specific direction on an image and alarm for a wrong way person trying to get into a sensitive area or show that when someone presents their smart card to a card reader that only one person walks through a doorway otherwise an alarm is triggered. Digital fences allow us to be alerted when someone crosses a boundary preset in the camera's recorded image.
All of this provides clearer videos and takes a lot of manual staring a screen away from weary operators who can typically only monitor 32 images for a maximum of 45min before missing significant events. Video analytics when programmed properly can reduce that fatigue and it doesn't need to go to the bathroom (insert Hollywood bank robbery storyline here). All of these commercial video security systems require Server and network design for bandwidth, image processing, and storage (which may be Govt. required minimums). Some sophisticated owners backed up in the cloud during late night data transfers to help reduce storage needs (this is sometimes called warm or cold storage) since you may not need to look at images more than 30 days ago, but you want to keep the archived footage for a year for example. I hope this enlightened you to how Security cameras and your favorite Netflix binge.
Randy Gruberman PE, RCDD, CxA
FL Southeast Market Leader/Low Voltage Service Leader