Insufficient Video quality to read a license plate

[PNelson]

I was unable to update my unreliability thread that I started earlier, so I created this new thread.

Last week, my neighbor had their garage broken into and they asked me to review my camera recordings for any suspicious activity.  I found a car that went down the alley three times within a 15 minute period.   The quality of the video was insufficient to read the front license plate from either the cloud or local copy.  The conditions were favorable as it was daylight, no glare, the camera is plugged into an outlet, and my signal strength for that camera is 2 out of 3 bars.  I did notice part of the footage was very clear, but other areas of the image like the license plate area were pixelated.  Due to video compression?   Possible Wifi interference?

I continue to have reliability problems when activity occurs, but the the camera fails to record during some activity occurs. 

The latency to view images in real time has gotten worse since 2019 when I bought the system.  To those those thinking they can remotely view and speak to someone at when they arrive at their front door, it is highly unlikely this will be possible by the time you get the notification and then attempt to activate the camera.   My internet service is 200 MBps up and down, so I should have sufficient speed.

When the batteries for the few cameras that I use on battery power die, or a camera fail, I will be get a hardwire system from an established security camera system that has a reputation for great customer service.   Or it might happen sooner because with more people out of work, more people going down the alley looking for things to collect or steal.   It will be more expensive to purchase and install, but at least it will have a better chance to be reliable,   For now, I hope the fact there is a visible camera will be a deterrent. 

Paul

[jguerdat]

There are a few issues for why a license plate can't be read. First, 1080p is only 2 megapixels and each pixel is covering a wide area given that the FOV is 140 degrees. That's roughly equivalent to an 8-10mm fisheye lens on a full frame DSLR. Download a high quality fisheye image and resample it to 2 MP and try to zoom in for details. Then add compression and all bets are off trying to get details at any distance.

Remember that these aren't ID cameras which would require a longer focal length/narrower FOV at the very least. These are activity cameras, showing a very wide angle  view. The Pro 3 and Ultra cameras, having 2k and 4k video respectively, could be better but then the FOV is also wider yet.

[PNelson]

---

@jguerdat wrote:

There are a few issues for why a license plate can't be read. First, 1080p is only 2 megapixels and each pixel is covering a wide area given that the FOV is 140 degrees. That's roughly equivalent to an 8-10mm fisheye lens on a full frame DSLR. Download a high quality fisheye image and resample it to 2 MP and try to zoom in for details. Then add compression and all bets are off trying to get details at any distance.

Remember that these aren't ID cameras which would require a longer focal length/narrower FOV at the very least. These are activity cameras, showing a very wide angle  view. The Pro 3 and Ultra cameras, having 2k and 4k video respectively, could be better but then the FOV is also wider yet.

---

Thanks for the clarification on the technical capability of the Pro2 and what it is feasible given the resolution.  My TV advertises 1080p which is rather detailed, but 2 megapixels is rather low by today's digital standards.   It would be unfair to criticize Arlo if the equipment is providing its advertised capability.   The realized quality can vary quite a bit depending on the compression that is used.   

I have had trouble with the reliability of this system and dealing with Arlo's poor customer service, I probably rushed to criticize.  

I think the thread is useful for someone trying to understand what the Pro2 is capable of providing and setting their expectations.   Your point that these camera are best suited to detect activity and not identification is an important point in setting customer expectations.