215-947-2338 (local)

866-633-3813 (toll-free)

That would be what we call a rhetorical question, because you can already tell by the fact that it is the title, that the answer will be No. This article explains the differences you need to know to make an intelligent buying decision and get the best solution for your needs.

Let’s define video analytics by saying what it is not. It is not a video camera with a simple motion detector. Several decades ago, when everyone got on the bandwagon of having video surveillance cameras everywhere, people figured out that no one really can watch the cameras. It is too boring. Human guards “tune out” after about fifteen minutes and they certainly can’t watch thirty cameras at once. No one looks.

So the security industry evolved putting motion detectors on cameras. Now the security company salesperson could walk into the field of view of your camera and it would immediately send an alert. Indoors, where nothing normally moves, that will do the trick. It will let you know when a thief is there. But outdoors, everything is always moving. Trees blow in the wind, leaves flutter, litter blows along, birds fly, shadows, sunbeams, and headlights all make changes in pixels on the sensor, which is all a motion detector knows. So hundreds, even thousands of false alerts per day.

To answer this, video analytics was created, utilizing machine-vision. Today nearly all video surveillance cameras offer “video analytics.” Reason for the quotation marks? It’s because sometimes the label is falsely used even though it is just the older type motion detection, a little bit refined to be able to rule out objects too small. The analytics is the software. Some is self-calibrating and is actually a neural network, capable of continuous learning so that its ability to classify objects and events constantly improves. True analytics recognize the human form and vehicles, based on numerous characteristics of how they look and move in, for example, a half million instances. Algorithms in the software compare a detected object with this model.

Some video analytic cameras provide “edge architecture.” This means that the processing is done in a hard drive right in the camera. If one camera fails it will not disrupt the whole system. Existing regular video cameras can be connected to what is called an encoder or analytic bridge and this will make them analytic, as well as serving as a local storage of data. Only when there is an alert will it be sent over the internet to your mobile device, or to your security staff or central monitoring service. This greatly reduces bandwidth, almost to zero, whereas former systems gobbled up bandwidth for offsite monitoring.

By the Rules

In rule based video analytics you can define “regions of interest” on the monitored view of each camera. For example, you care about what happens in your storage yard, but not out on the public sidewalk or street. Thus a virtual fence. You define times of day. The rule may be that no one is allowed there. Or it may be that vehicles can only drive one direction but not another across a “directional line.” There are rules for object left behind or object removed. The great thing is that manufacturers like Avigilon have 8 and 16 megapixel cameras with analytics. This means that a single camera can cover a vast area of space with high resolution. Higher millimeter lens, although narrowing the angle of view, can analytically detect an intruder a mile away.

Other rules are for loitering or dwell time. There are rules for “more than X number of people in a defined region” — to possibly detect crowds which could form around a fight in a schoolyard, for example. Special applications allow counting of people so you could tell if everyone exited a building in a fire drill.

Technical Stuff

There is a whole world of technical considerations that determines the quality of the image capture. Use of gigabit Ethernet is 100 times faster than 10 megabit data transmission. High image rate, lossless compression, high quality recording, image enhancement, dynamic range to see where part of image is bright and part is dark and where the subject moves across varying areas of light, scientific grade sensors which are three times as sensitive as standard commercial ones capture details of moving objects, linearity, intelligent adjustment of image parameters. My head is spinning! But you understand that things vary greatly between manufacturers.

My simple suggestion is to compare and look for where something doesn’t work, not for where the salesperson directs your attention to where it does work. Someone running in the distance in low light. Will it detect them?