Member Discussion

Analytics For Residential Executive Protection

Can anyone suggest an intruder detection type analytic that would highlight a potential intruder in the video stream day or night? This is for a residential type application, if the customer hears something outside they would like to pull up all the cameras and quickly see where the problem may be. Typically the angle of the camera will be in the soffit of the house looking across the side to view windows and doors. We are using Exacqvision for the VMS and deciding on either Axis P3364-LVE or Samsung SNV-7084R.

Axis's tripwire app worked well in our test and your choice of an integrated IR camera will help in the one key difficulty we found (low light).

Also, Samsung's VMD worked well in our tests.

I'd start with considering those 2 as they are also the least expensive and complex ones to add.

John, do you know if either of these will highlight or put some kind of border around the moving person or object?

Axis does not, but Samsung will embed a green bounding box around whatever is triggering its VMD. We talk about it at around 2:50 here:

...if the customer hears something outside

Sounds like in addition to VMD you may want some type of rudimentary audio analytic to determine what camera to look at when there is no video event.  In my personal experience, having the minimum capability of being able to easily replay audio triggers and to know at least what cameras to look at is key to peace of mind, esp. in the absence of video triggers. Exacqvision 5.8 apparently has some audio searching capabilities shown here: 

If you need more capabilities, actual audio analytics (window break, aggresion detection) are discussed here, also Axis cameras apparently can run audio analytics on the camera side, though I have not used them personally. 

Thank you for the suggestion an audio analytic would be a nice feature to look into for this. I will look into discussions.

Be very careful about audio analytics, especially if you are using built in microphones. Even if you add in microphones, audio analytics are not a mature area and you should be careful about introducing problematic false alarms.

Agree with John about analytics and the false alarms. Where a certain degree of false alarms might be acceptable in a 24/7 environment, in a residential setting, after the second or third time you have a 3 AM false alarm they'll want it shut off, and you might not want to spend the time fine tuning it to work better. But there are simpler ways.

Expanding on what I said before, if the VMS or the camera can simply do loudness detection, (the audio equivalent of VMD), and generate non-audible events which log the time and the camera (and the db level), that is invaluable when trying to determine "what/when/where was that?" in the middle of the night, and even more so the next morning!  And if there are a few extra markers in the stream its easy enough to reject them at that time.  For extra points you can speedup the framerate for a bit, on the event (if supported by VMS), or trigger a light through camera I/O.

Some Samsung cameras, like this WisenetIII SNB-5004, offer such basic detection built-in, that can trigger VMS events (if the cameras alarm i/o is supported by the VMS), shown here:

I would recommend using some good mics from sennheiser, eg ME-102. They are a little bit expensive but quality is key.

.......and then he runs outside to protect his home and they own him......

A VideoIQ Rialto unit, or R1 series, can be used with just about any camera you choose. It will identify movement and create an alert, then send a short preview clip of whatever caused the event. It can distinguish between vehicles and people and does a pretty good job of eliminating false alerts from animals and moving branches etc.

That might work for you.

Beyond the fact that Avigilon has just end of life'd VideoIQ Rialto, the price of adding one of those units dramatically increases the cost of the project.

Alan, if you consider that, you will have to weigh the price increase (thousands more) to the potential accuracy benefits.

I'd recommend trying analytics on your cameras first, seeing how accurate they are and then, if you have problems, considering something like the Rialto.

Yes, the cost of the tech is 2k to 2.5k per 2 to 4 cameras depending on the resolution you're going to run them at. Given he was speaking of executive protection, I wasn't sure how much of a factor price would really be. The analytics work well, and that seemed to be the focus of the question.

As to the end of life issue, accoring to our conversations with Avigilon this morning, the encoders will still be available in some re-branded form. The cameras are going to go away, but the encoders will not we're told.

Depending on what you're trying to accomplish, it may not be your best option. It is an option though.

The encoders will still be available in some re-branded form...

An unsung IPVM member saw it coming the day of the acquisition.