The Farsight experiment: Putting video analytics to the test

Effective Surveillance with video analytics

Use of video analytics is on the rise. More and more customers are looking for intelligent security solutions that automatically detect unusual activity.

As a result, more analytic solutions have become available on the market – and they claim to offer very different results for the customer.

To get to the bottom of the pros and cons of each solution, the security experts at Farsight have teamed up with four analytic providers to put them to a head-to-head test.

Objectives:

We want to get under the skin of analytic technology and establish:

  • Do all video analytic solutions provide the same result, for example reducing false alarms and providing accurate alarm information?
  • Which solutions offer the best value for money?
  • Could a free video analytics solution perform as well as one costing thousands of pounds?
  • Do the solutions fulfill the promises they make?

Method:

We have carefully selected four partners – each from different price points in the marketplace. As a result, the four solutions we have chosen to test range from free to thousands of pounds.

Four 720p Axis IP cameras will be fixed to the side of the Farsight Observatory. All four cameras will look at the same scene with a height different of approximately 50cm between the lowest and highest cameras. This ensures the cameras are as close as possible to being at the same height as each other, whilst presenting the same viewing angle.

To make the test as tricky as possible, we’ve thrown in a couple of curveballs to really challenge the video analytic solutions:

  • 50% of the image is tree and sky, the other 50% is ground (the Farsight closed compound car park),
  • and shrouds have not been installed to keep rain off the lens covers.

Over a six week period we will be running several tests. In each of these tests the video analytics solutions will be assessed on their ability to perform in a different scenario.

For each test:

  • each video analytics solution will be configured with the same scenario,
  • the alarms from each solution will be counted and recorded daily,
  • weather conditions and observational notes will also be recorded daily.

So what scenarios do we intend to test the video analytics under?

  1. The lazy person test: the whole screen is a detection zone for cars and people.
  2. The targeted zone test: unwanted areas are masked out and only the car park area is a detection zone for cars and people.
  3. The trip wire test: a virtual directional trip wire is created across the car park.
  4. The loitering test: an area is created where a time delay is built in before an activation is sent.
  5. The combination test: areas and trip wires will be combined for a more complex scene.

What happens next?

The cameras are currently in place with the different video analytics solutions up and running. We’ll be rolling out the different scenarios over the next six weeks and analysing the results soon afterwards.

Watch this space for the full report on the results and our experts’ thoughts on how each solution performed.

You can be amongst the first to receive updates on the tests and receive the full report by signing up to our newsletters in the bottom right-hand corner of this page.

Make sure you stay tuned: later this year we will be running tests that combine PIR technology and analytics – will these two security solutions work together to reduce false alarms even further?