Hikvision camera smart event Intrusion Detection

[Martin Perks]

Hi

Can anyone confirm when setting up a camera on a DVR/NVR and using smart event ( Intrusion Detection ) should the camera image be set to a certain resolution, also should the camera only be viewing a specific area rather than a wide overview image.
We have noticed pixelation of the recorded image when showing an alarm ( mainly false alarms) as if the analytics are using 4:3 ratio but the image is a 16:9.

False alarms are a continual problem across multple sites and it appears of if the analytics are better suited to a more defined ( tighter) image, example, as in an image seen through a 6-8mm lens rather than a wider view a 3mm lens would give.

We also have repeat activations when it rains no matter what settings we use.

Any help would be appreciated.

 

[JB1970]

We have noticed pixelation of the recorded image when showing an alarm ( mainly false alarms) as if the analytics are using 4:3 ratio but the image is a 16:9.

This often happens if you have different settings for the main stream between normal and event. The NVR records the main stream (normal) settings continually but on alarm trigger the main stream (event) parameters are used. If the bit rate for event is less than the bit rate for normal you may see degradation during the event capture. As it's the NVR that makes that stream adjustment, the setting is in the NVR video parameters for each camera.

Other possible causes:
- too low a bit rate when using Variable Bit Rate. A bit rate that's set too low may be adequate when there's a relatively static image but can be poor during periods of motion as the compression increases in an attempt to keep within the configured maximum bit rate.

- variable bit rate failing to respond quickly enough and ramp up the bit rate in response to the motion in the image

We also have repeat activations when it rains no matter what settings we use.

I'd expect that if you're not using AcuSense cameras and filtering the events for Human/Vehicle. Regardless of sensitivity and configuration settings it's difficult to filter out unwanted alerts.

 

[Martin Perks]

Hi Thank you for the reply, if anything we have the event recording at a higher rate than the normal but in most cases at the moment they are set the same. We generally use our own networks rather than join a customers so the bit rate is set to the cameras requirements but we have increased the bit rate but with no real advantage, even if the cameras goes straight to the recorder we still experience false alarms. I do appreciate lighting, position, network, model all make a difference but as a general view do you think a 120 degree view ( which gives a slight curve to the image) or a 90 degree view ( providing straighter lines) of the same image would give more accurate analytics for Human/Vehicle events.