Simple Analytics .... Intrusion Detection

Hi everyone, I'm about to install a system, ~30 cameras in total. The cameras are to be temporary secured onto the side of some wooden hoarding - whose total perimeter is ~600 metres.

The cameras are to go apx every 30 metres and pointing then in from the hoarding at an angle of ~30 to 40 degrees. In this way ...we're someone to climb over the hoarding and into the site the cameras video analytics would pick them up.

I've always used analogue cameras combined with redwall 3020 pir detectors connected to a dvr made a company called Teleeye. I would continue to do the same were the site apx 3 times smaller.

So I'm going down the ip route.

I have in mind to buy a load of maybe Hikvision cameras which apparently are able to see 50m in the dark and do have some limited Analytics on board. The model number is DS-2CD2232-I5.

The camera is a 3mp bullet type. Should the analytics be no good it also comes with an alarm input - so could install them with PIRs if needed

I'm not looking for particularly advanced analytics.

The site concerned is a construction site. The camera simply has to pick people up within 30 metres or so of it. Ideally it will ignore foxes...but no great expectations on that front.

An alternative purchase is the Hikvision DS-2CD4A26FWD-IZHS. Same make of camera but triple the cost and perhaps with more effective analytics.

I hope to be able to "get away" with the first option.

I've been assured it will reach 50 metres but I'd be perfectly happy if it made 30m. If it could some how pick up anyone walking past it and with less false alarms than mere motion detection I'd be happy. I'm not after left baggage alert or counting people etc....but the ability to register and pick up a human.

I'll also be using a 32 way hikvision nvr connected to a monitoring station using either the sentinal or immix Platform.

Any suggestions most welcome.


Will you elaborate on what you mean by "the ability to register and pick up a human"?

Do you mean reconize the movement is not a dog, but a person? or something else?

In case you haven't seen, this is a video from 2013 showing the Hikvision Intrusion Analytic running on an unspecified model Hik. I'm assuming their newest analytics are at least as good.

Note the scene with the cop and the dog that the alarm does not trigger on the dog.

Still though, anything automatically calling central station would have an extremely low false rate and/or video verification to be viable.

I'm not confident that a single trigger system in an open outdoor space can be configured to have both a low number of missed events *and* and low number of false alerts at the same time. See this discussion.

Are you seriously judging video analytic performance based on a vendor marketing video???

Ethan finished a test of Hikvision video analytics last week. The report will be released later this month.

Besides assuming that Hik analytics haven't gotten any worse in two years, I'm not judging the video at all actually. Only stating the fact that the dog doesn't trigger the alarm, in the clip.

If I were considering buying Hikvision intrusion detection and hadn't seen this video, I would want to see it, wouldn't you?

I looked at 10 or so videos and this was by far the most varied and least produced. And I'm sure any embarrassing triggers or misses would have been edited, but I doubt it's a straight forgery. And the length of the continuous clips is better than the usual PR BS.

Also, the OP's requirement was pretty basic "if it could some how pick up anyone walking past it and with less false alarms than mere motion detection I'd be happy."

In any event, I ended skeptically mainly because of the central station hookup.

Are these vms/nvr analytics? The only edge analytics I saw for hikvision at ISC West was from agenvi.

No, Hikvision is offering their own suite of embedded analytics inside their cameras.

Thanks for replying Brian.

To tell the difference between a human and a dog ...would be nice

Although where I'm foxes / cats that are problem - way more so than dogs.

I have a store room full of redwall pir external motion detectors and genesis detectors (made by luminite ) ...but in going digital I'd like to do away with them eventually.

These detectors pick up / detect humans .... as well as cats and foxes.

Motion detection on cameras picks up just about everything ....including trees and grass blowing in the wind car head lights etc.

Between motion detection and infra red pirs ...I chose the pirs

In terms of analytics ...where a square of some sort is drawn around the cause of alarm be it a human or cat , the operator at the monitoring station will be instantly drawn to the sours of activation.

If he sees the analytics has nicely outlined a bloody fox or cat or human ...then at least he knows the case instantly

I must I'd prefer the analytics to ignore the foxes and cats ... but hey ho - not essential....but would be nice

To recognise a man jumping over a fence or man walking staroght towards the camera and send an alert would be perfect.

No facial recognition of any kind needed.

Ideally tho camera would ignore grass blowing in the wind or car head lights....

There is nothing simle about changing from analog to IP and adding analytics.

Glenn - I am from Hikvision. The difference in the analytics in the 2-series vs the 4-series are mainly a matter scaling. The 2-series support one zone and one line where the 4-series supports 4 zones and 4 lines, in addition to other analytics functions. Both will not identify/classify subjects so it is a matter of tweaking the sensitivity. The 2-series is positioned as having "good" analtics and the 4-series as having "better" analytics, but I do not have quantitative data to support these classifications. I suggest doing some testing with the 2232 to see how well it will perform in your application.

I think you have set reasonable expectations and the selection of EXIR bullets for the better IR coverage is a good choice.

For line crossing / tripwire, is the analytic performance / algorithms the same on both the 1 and the 4 series. In other words, outside of the more zones and lines on the 4 series, should performance be expected to be similar for the 2 series?