Fps and recognition

[aurmol 0]

Is there a minimum frame per second before it is useful in security?

 

Im using a 5mp with 12 fps.. and I can't seem to recognize people who move so fast and it appears above certain speed, you only see a blur.. What is the consensus on this? Thiefs don't stop at cctv and pause.. they even move faster... is 30 fps per second a minimum for recognition? or 20 fps? What? is it same speed at low light and daytime? What is your experience on this? Thanks.

[ssmith10pn 0]

Maybe it's the camera?

I get perfect images at 5fps.

 

What's your key frame interval?

[aurmol 0]

Maybe it's the camera?

I get perfect images at 5fps.

 

What's your key frame interval?

 

24 key frame interval?

 

Maybe it's the exposure? What is your exposure value? What is the minimum where a person face can't move fast (even an athlete) past a certain milliseconds?

 

In those 30fps versus 20fps vs 10fps video. So they all can be certain same exposure and the frame per second is just the sampling of those same exposure?

[Don Stephens 0]

Your cameras shutter speed, which they all have but is only disclosed with LPR/LPC cameras, is what's ultimately going to determine what you can capture in detail at various speeds. If it hasn't clicked yet, that statement isn't going to help anyone with their own personal setup as it's going to be camera dependent. For the most part, you're just going to have to try it and see what happens.

 

I've been easily able to achieve facial recognition as low as 3fps with IP cameras for anyone moving at a casual pace. It's also hugely going to depend on how long they're within the cameras FOV.

[aurmol 0]

Your cameras shutter speed, which they all have but is only disclosed with LPR/LPC cameras, is what's ultimately going to determine what you can capture in detail at various speeds. If it hasn't clicked yet, that statement isn't going to help anyone with their own personal setup as it's going to be camera dependent. For the most part, you're just going to have to try it and see what happens.

 

I've been easily able to achieve facial recognition as low as 3fps with IP cameras for anyone moving at a casual pace. It's also hugely going to depend on how long they're within the cameras FOV.

 

I've been testing all night the low light performance of the dahua 12fps 5Mp. With exposure set to auto, motion is mostly a blur.. I have to set Exposure to Low Motion Blur to get good facial recognition.. but with IR on, there is a lot of noise. I wonder if the 4300S would be a big improvement. I'll exchange the 5mp for two 3mp 4300s tomorrow and hope there would be major improvement or i'll stuck with two 3mp (which I can't return anymore).

 

Do you also set the 4300S to Low Motion Blur in the exposure at nighttime to get facial recognition or does Auto exposure work?

[Don Stephens 0]

I've been steered clear of that 5MP camera a couple times, but I didn't want to say anything since I've never tested it for myself. The overall opinion doesn't seem to be to good with that camera.

 

At the distances and lighting conditions I've run the 4300S in, I've been pretty happy just leaving it alone as far as the settings are concerned. I'm sure you'll find that some settings can be toggled to optimize it to your own environment (it's going to vary).

[thewireguys 3]

Your issue has nothing to do with FPS and everything to do with lighting, exposure and lensing.

[ssmith10pn 0]

This^^

[MaxIcon 0]

To avoid motion blur, you need to set the maximum exposure at 1/30 sec, or 33 ms, depending on how your cam sets it. This will be good for people walking or slow cars, but for running people or faster motion, you need 1/60 sec (17 ms). If you're on PAL, that may be 1/50 sec. The low motion blur setting probably uses 1/30 or 1/25 sec, but I like to set it manually to be sure.

 

The problem is that this limits the light it captures, and the image gets noisy unless you have good lighting or a large sensor with a fast lens, as the posters above mentioned.