Human Detection false positives on the C340
Hi,
My cameras often give me false positives (alarms) with human detection. I have enabled Active Defense - Human Detection, Camera Tempering. In the Human Detection section, I have set a detection area, The sensitivity is set to 50. These are the scenarios under which I get false alarms. These are when I get false Human detection alarms.
- Fibers/Spider web threads flutter in front of the camera in a breeze. This is not a huge spider web...just a single thread. It could also be just a thin twig.
- When a light is switched off and on again
- When a small insect crawls/flies across the lens
- When it is raining and leaves on a tree move
All of the above happen independently and in the night. This is also because I only arm the cameras at night.
Isn't the whole idea of human detection to have something that resembles a human to cause an alarm?
Another question:
Does Intrusion detection with Human classification use the same model/algorithm to identify a human? Is one superior to the other?
To use intrusion detection instead of human detection in the above scenarios, I need to identify a practical intrusion time and percentage. Since these insects are close to the lens they may take up about 20-30% of the frame at a given time. Humans at a distance will also occupy about the same frame percentage. Regardless shouldn't the human form be distinctly different from the above?
@Clive_A
My answers inline:
Clive_A wrote
Hi @vin32
Thanks for posting in our business forum.
Do you have other types of detection enabled?
>Human Detection and Camera Camera Tampering only.
Are you sure every notification/alarm is classified as Human Detection? Maybe they are Motion Detection. Not really from Human Detection?
>The notification clearly says Human Detected. I have attached a picture from my app.
The Human Detection would detect things with things with at least four limbs and a head.
An object similar to a human with four limbs would be considered a human and it may trigger the detection.
>tried to attach a video of the event but couldn't. the video has pretty much the same image as above except you will see the string/fiber swaying a little and a tiny insect (almost a spec) flying past. Nothing close to a human.
I have seen this happen with other cameras in the home routinely. It is particularly bad with detection in the rain. I have tried to draw boundaries for detection around trees but that hasn't helped either.
If necessary, please post some pictures of the notifications in your app.
I would also like an answer to my other question (attached below for reference)
>Does Intrusion detection with Human classification use the same model/algorithm to identify a human? Is one superior to the other?
To use intrusion detection instead of human detection in the above scenarios, I need to identify a practical intrusion time and percentage. Since these insects are close to the lens they may take up about 20-30% of the frame at a given time. Humans at a distance will also occupy about the same frame percentage. Regardless shouldn't the human form be distinctly different from the above?
Hi @vin32
Thanks for posting in our business forum.
vin32 wrote
@Clive_A
My answers inline:Clive_A wrote
Hi @vin32
Thanks for posting in our business forum.
Do you have other types of detection enabled?
>Human Detection and Camera Camera Tampering only.
Are you sure every notification/alarm is classified as Human Detection? Maybe they are Motion Detection. Not really from Human Detection?
>The notification clearly says Human Detected. I have attached a picture from my app.
The Human Detection would detect things with things with at least four limbs and a head.
An object similar to a human with four limbs would be considered a human and it may trigger the detection.
>tried to attach a video of the event but couldn't. the video has pretty much the same image as above except you will see the string/fiber swaying a little and a tiny insect (almost a spec) flying past. Nothing close to a human.
I have seen this happen with other cameras in the home routinely. It is particularly bad with detection in the rain. I have tried to draw boundaries for detection around trees but that hasn't helped either.
If necessary, please post some pictures of the notifications in your app.
I would also like an answer to my other question (attached below for reference)
>Does Intrusion detection with Human classification use the same model/algorithm to identify a human? Is one superior to the other?To use intrusion detection instead of human detection in the above scenarios, I need to identify a practical intrusion time and percentage. Since these insects are close to the lens they may take up about 20-30% of the frame at a given time. Humans at a distance will also occupy about the same frame percentage. Regardless shouldn't the human form be distinctly different from the above?
I need a link to the video. You may upload it to the public video platform.
About the questions, no.
If your model is C340 hardware V2, we may get a beta for you with the optimization of the detection mechanism.
Will send you an email and please reply to the email with the video link.