Statis object detection

[iku777]

Hi,

I need some help/guidance in configuring properly static object detection.
Here is my issue. I have a tree and when it winds, it generates shadows / movements, which trigger my motion detection.
Given I have some AI that triggers only alerts on car / people, that is not a problem as per say.
Problem is that quite often, it generates a warning for the car we have parked all the time in the driveway

I have enabled the static object analysis (see AI picture). I have attached my settings as well as the details of the rule I have (Static object and Rule picture).
I have read the manual and configured those options to my undestanding, but somehow it still generates a ton of false alerts.
Am I missing something?

Thanks for your advices.

[TimG]

Does anybody have a clue ? I'm having similar problems so if you do understand this, please post

[BruceH]

I am no expert, but why do you have static selected, it is my understanding that is to identify static object (ie not moving), is it not the movement of the trees that is the issue. What exactly are you trying to alert on from that camera, do you really need to detect the non moving car?

[iku777]

I am no expert, but why do you have static selected, it is my understanding that is to identify static object (ie not moving), is it not the movement of the trees that is the issue. What exactly are you trying to alert on from that camera, do you really need to detect the non moving car?

Hi,
Happy to explain. I don't want any alerts on movement of trees / shadows etc. I have programmed Blue Iris to only triggers alerts on motion detection where either a person or a car is detected. When my car is not parked there "permanently', this works fine. A movement of the trees and it's shadow would trigger a motion detected that would get ignore as no car / persons are detected.

Problem I have is that, when the tree /shadow creates a movement, the motion detection gets triggered and sometimes BI identifies the car, even though it didnt move for the last few days.

Desired outcome would be: motion detected, car detected, alert ignored given that car was there and hasn't moved in the last x mins / hours.

Does that clarify?

[Pogo]

Mask the location of the car's 'static' location. It then becomes a non-issue unless it either exits or approaches the masked area which would presumably trigger the appropriate action for the detected activity.

[iku777]

Mask the location of the car's 'static' location. It then becomes a non-issue unless it either exits or approaches the masked area which would presumably trigger the appropriate action for the detected activity.

Hi.

Yes, this seems to be an idea to get around the problem.
Assuming the 'static' car parks at a very similar location everytime, I don't see a lot of downside to this option.
Just thought it was basically the purpose of the static object funtionality and that there was a way to configure it 'better' to have it fit my need.

[Pogo]

Just thought it was basically the purpose of the static object funtionality and that there was a way to configure it 'better' to have it fit my need.

There probably is, but intuitive or intelligent doesn't seem to apply to how AI is actually designed. LOL

[TimG]

I'm certain its purpose is intelligent masking rather than simple black out masking. One of my issues is reflected cloud movement on my car windscreen causing multiple CPAI "Car" triggers. Surely it can't be impossible to show the car isn't moving

I've just set up a rule. It didn't have any, so no rules = no static objects. This rule looks to be the opposite to what I'm after, so I'll try this for laughs:

static.png (65.44 KiB) Viewed 161 times

[TimG]

Hmm, I told myself to RTFM, and it has this rule that looks somehow useful:

static2.png (38.2 KiB) Viewed 158 times

If only I knew what it meant

[lightmaster]

This seems like a serious issue to me, at least in my case causing so many false positive alerts by AI finding a neighbor's light to look like a car, or whatever and setting off so many notifications that you stop even looking at the notifications, which defeats the purpose of having AI or even having cameras to begin with.

I'm actually thinking there is a bug involved here. I have my driveway and front yard camera set to check for static cars, and set a static car as a "Canceled Alert". It does indeed show up in my alerts list as a "Canceled Alert" (meaning I can right-click the alert and have the option to "confirm alert", whereas a normal alert with a moving car gives me the option to "cancel alert" when I right click it). Problem is that this "Canceled Alert" shows up in the Alerts folder and not in the "canceled alerts" folder in the windows on the left side, and it still triggers an alert trigger, sends a notification to my phone, and keeps the main stream recording for this time frame and not the sub-stream recording (wasting HDD space). It seems logical to me that while it is correctly being set as a "Canceled Alert", it's incorrectly being put into the wrong alerts folder.

Also, I did try setting a memo of "static-car" and adding that to the "To cancel" list, but then that means that when a person is walking around and a static car is present, the alert is still canceled instead of correctly being alerted due to the non-static person in frame. I also tried using "Occupied Alert", but that resulted in the same outcome as "Canceled Alert", except it has the occupied icon instead of the canceled icon. Prior to whichever update mucked up the static object analysis from a simple checkbox, "Occupied Alerts" would be sent to the canceled alerts list and never set off any notifications.

Mask the location of the car's 'static' location. It then becomes a non-issue unless it either exits or approaches the masked area which would presumably trigger the appropriate action for the detected activity.

Masking out an entire area because of an object that's sometimes static there is 100% not the correct approach. There are plenty of scenarios where you need to be able to get alerts from in front of a static object, but not alerts by the static object itself. A mask means there's an entire volume that stretches from the lens of your camera to infinity where AI can't see anything, not that there's a small area where the static object is that's hidden. Masking out may be appropriate for areas above the horizon where there's nothing but trees to needlessly set off motion and you aren't anticipating Tarzan to come swinging into your yard, but not for something that's at ground level and will obscure areas where notification is actually needed.