Hikvison G5 cameras - Human intrusion detection false alarms?

[LoinKing]

I have a few G5 ColorVu cameras (Dome and Turret) with latest firmwares. I've set up Intrusion detection on them, checked Human and set Target validity to 'Higher'. Still I'm getting false alarms because of cats at night. Interestingly, earlier I had Motion Detection setup with Human checked, but never had false alarms.

Why is there detection accuracy difference between Motion and Intrusion detection? Is there a way to fix it? I don't want to use Highest setting to avoid missing on actual human alerts.

 

[Kyle]

From our experience, intrusion detection on the Acusense cameras is generally pretty accurate (although someone here may put me right, as we only have experience testing the cameras in one location).

Are the cameras connected to an NVR, and is this where you are setting up the events if so?

What are the model numbers of everything, and running which firmware?
I know the event performance has been tweaked through firmware upgrades on these cameras.

Are you drawing the maximum and minimum target size boxes as well?

 

[alfaclvi]

I have similar issues from the beggining and initial firmwares.
For example 2-4 meters from camera I draw region area for Motion and on exact same area I draw Intrusion Detection.
Motion detection on 20 or 40 detects every kind and every size of target inside this area Human or not.
In same area Intrusion Detection some times does not detect target at minimum and max size(not set), at lowest validity, at max 100 "sensitivity"(percentage of target size), at lowest validity(basic). And I am not talking about insects but for Humans.
Detection for Smart events should always exist and this is not a matter of Sensitivity or target size or validity. The target must always be Detected(its the D on DORI), processed(Observed, Recognized or Identified) and give Alarm according to our Settings and what we want to give an Alarm.
For example the target will give an alarm after loittering for 5 secongs inside the area but the Detection and Rectangle surrounding the target should exist 5 seconds before Alarm.
Possibly this issue has to do with the way of movement. Straight line ahead, sideways, diagonal, making rounds. An insect for example flying around is Detected. A Human quickly(not running) passing by this area is not detected. Possibly the reason is the position of the camera or the light, sun or other, in the area. But this cannot be an excuse because the Motion(Basic events) Detects. So this is definitely the firmware. The very strange thing is that the far targets at 10 meters distance are always detected. Nothing is missed.

And yes. Cats and sparrows many many times are detected as Humans.

 

[alfaclvi]

What I had noticed when I installed the 5.7.17 on G5 was that this was more accurate. Less sensitivity but more accurate on Human detection. No birds and no cats at all. I don't know if many people want this, but many others may not. I would just like everything to be adjustable according to my needs.
On 5.7.17 Motion has an issue, at least for me and my NXI-K NVR, for the time being, so I wont use this till they fix. Others possibly M series have not this kind of issue. So there is no rush to Update. Just to experiment

 

[LoinKing]

On 5.7.17 Motion has an issue, at least for me and my NXI-K NVR

What issue did you experience with motion? I also have NXI-K nvr and updated to latest firmware today.

 

[alfaclvi]

@LoinKing

Firmware - New firmware V5.7.17build 240424 for ColorVu G5

Updated on 5.7.17_240606 Motion Detection is not working at all on DS-2CD2047G2-L and Acusense DS-7608NXI-K2(V4.82.100build 240606) with previous settings Increased Sensitvity to 100, there is Detection, but motion Event is not uploaded as VCA. Enable Dynamic Analysis for Motion is now renamed...

www.use-ip.co.uk

I can't solve this as I mentioned above

 

[alfaclvi]

I also have NXI-K nvr and updated to latest firmware today.

Also cannot go back(roll back) from this new firmware 4.82.100. And they do not mention on Release Notes that appeared 13 days later !!!!!

 

[JB1970]

Why is there detection accuracy difference between Motion and Intrusion detection? Is there a way to fix it? I don't want to use Highest setting to avoid missing on actual human alerts.

That's the million dollar question...With AcuSense, Motion used to (seem to) be the least accurate, while Intrusion and Line were far more reliable due to the additional parameters that are set (Minimum Size, Maximum Size, Sensitivity, Direction [Line], and Target Validity). More recently, sometimes I find that motion results in less false alerts. The sensitivity for motion is complete BS. Allegedly setting a lower sensitivity, a trigger will only occur for larger targets...I only ever set sensitivity at 20, and even then an 8MP G5 platform camera with a 2.8mm lens triggers an event when a Human appears on screen (roughly 45 metres away and occupying 5% screen height). However, I get less alerts from cats, foxes, pigeons, etc on Motion than I do on Line and Intrusion.

What Hik should implement (IMHO), is the ability to set different trigger actions for different motion, line, intrusion areas. I'd like to set 1 motion area to only bookmark the video, with a second triggering a notification ('Notify Surveillance Center'). Currently I need to enable motion full screen to bookmark the video and select an alternative VCA (line, intrusion) as the trigger for a notification. Allowing different trigger actions for different areas using the same detection would offer way more flexibility in setup.