What does it take to truly identify individuals at night?

[jlppilot]

I have a thread in the networking section where I'm trying to set up a system that involves transfering data from a bunch of cameras 1.4 miles over the LAN. It's possible I would have as many as 30 cameras someday in the location that is 1.4 miles away.

I'm trying to decide how fast that 1.4 mile connection needs to be. I want to be able to truly identifiy people who visit that location at night. What kind of cameras will I need to do that and what kind of bandwidth will they need?

From what I have read, high resolution cameras don't work great in low-light conditions. However, low resolution cameras aren't great for identifying people.

So I guess where I need to start is, "what resolution is required to truly identify people at say 50 feet away?" What if I want to be able to truly read a license plate at that distance or maybe even 75 feet?

After the resolution is decided, then I just need to install the cameras and see how things look and then start increasing lighting as required until things work OK.

For those interested in more details about the entire setup you can see more here:

Long range system

I have two farmsteads and each farmstead has multiple buildings I need to monitor as partially described in the picture. Location B actually had 6 different areas I would like to monitor. Each area needs between 2 and 6 cameras. Location A actually has 5 areas I would like to monitor. Each...

ipcamtalk.com

If you want the short version, I have two farmsteads that I want to monitor. They are 1.4 miles from each other and I'm 15 minutes from the nearest town and I have had some stuff stolen and I'm not interested in being a helpless victim anymore. I want to increase lighting and surveillance on both farmsteads in such a way that makes it so the evidence I collect allows law enforcement to rapidly capture offenders. My biggest problem is lack of good internet in my remote location and lack of ethernet cables being pulled to each building so I'm going to have to build a robust wireless LAN and then rely on cellular internet for viewing alerts that the Blue Iris system sends.

 

[mat200]

...

From what I have read, high resolution cameras don't work great in low-light conditions. However, low resolution cameras aren't great for identifying people.

So I guess where I need to start is, "what resolution is required to truly identify people at say 50 feet away?" What if I want to be able to truly read a license plate at that distance or maybe even 75 feet?
..

Hi @jlppilot

Q: "what resolution is required to truly identify people at say 50 feet away? What if I want to be able to truly read a license plate at that distance or maybe even 75 feet?"

A: So it really is a question of

• Effective pixels on target
• and a good angel on the suspect to get a good facial ID image.

100 ppf is the working number we like to use ( it is based on various studies )
6-8 feet high is the typical install height of the camera we recommend ( this allows a good angel for most homes - of course as you have a farm you may want to capture images further away from the camera position.. )

See the DORI section in the cliff notes
See the IPVM calculator

 

[sebastiantombs]

From the layouts you provided in the other thread, you're going to need a boat load of Nano Station Loco equipment. That will also provide the link between sites although you may want to invest in a more robust version than a Nano.

It's no so much a question of resolution, it's a question of pixels at a specific distance that make identification possible. That is usually more dependent on the focal length of the lens than anything else. Too often people make the mistake of trying to cover a wide area with a single camera. That does provide a nice overview but the short focal length makes any meaningful identification impossible, too few pixels at anything over, say 10 to 20 feet or so. The other trick is to mount the cameras relatively low so you're not just getting the tops of their heads and shoulders. Six to eight feet is about maximum.

Bandwidth, if you use a BI machine at the remote end, will be relatively limited by BI when viewing all cameras at once, but when you look at a single camera or clip the bandwidth will be sufficient for good resolution.

 

[jlppilot]

From the layouts you provided in the other thread, you're going to need a boat load of Nano Station Loco equipment. That will also provide the link between sites although you may want to invest in a more robust version than a Nano.

Thanks. I would love any suggestions on equipment there. I enjoy technical things but this is not my field of expertise so this will be a steep learning curve for me. I really don't want to spend more than $10,000 the first year so this is not something I'll be able to do all in one step.

It's no so much a question of resolution, it's a question of pixels at a specific distance that make identification possible. That is usually more dependent on the focal length of the lens than anything else. Too often people make the mistake of trying to cover a wide area with a single camera. That does provide a nice overview but the short focal length makes any meaningful identification impossible, too few pixels at anything over, say 10 to 20 feet or so. The other trick is to mount the cameras relatively low so you're not just getting the tops of their heads and shoulders. Six to eight feet is about maximum.

That makes sense to me and is one of the reasons I have so many cameras. Do you have suggestions on cameras that can identify people or license plates in the 50 to 75 foot range?

Bandwidth, if you use a BI machine at the remote end, will be relatively limited by BI when viewing all cameras at once, but when you look at a single camera or clip the bandwidth will be sufficient for good resolution.

Can you elaborate on this part? By that I mean can you describe the specific limitations of Blue Iris software? I'm worried about the kind of computer I'll eventually need to run this system. I would think maybe I'll need 64 gigs of RAM and I can't imagine how much processing capability? I guess I assumed that the Blue Iris machine would always have 100% of all of the cameras data in RAM for the last 20 to 40 seconds and would dump anything any older than that. If a specific camera detected a motion event I assumed Blue Iris would go back 20 or 40 seconds and start recording that to the hard drive and would send an alert to your phone allowing you to view that camera over the Internet if you so desired.

I assumed that any camera you focused on would have very good resolution available to it and any time it recorded to the hard drive it would record full resolution as well. Maybe that's not how it works.

 

[jlppilot]

Hi @jlppilot

Q: "what resolution is required to truly identify people at say 50 feet away? What if I want to be able to truly read a license plate at that distance or maybe even 75 feet?"

A: So it really is a question of

• Effective pixels on target
• and a good angel on the suspect to get a good facial ID image.

100 ppf is the working number we like to use ( it is based on various studies )
6-8 feet high is the typical install height of the camera we recommend ( this allows a good angel for most homes - of course as you have a farm you may want to capture images further away from the camera position.. )

See the DORI section in the cliff notes
See the IPVM calculator

Thanks. I I will look through the DORI section of the cliff notes and the IPVM calculator. The 100 ppf is good information. I can tell this is already going to help me refine what I need to do a lot.

 

[mat200]

Thanks. I would love any suggestions on equipment there. I enjoy technical things but this is not my field of expertise so this will be a steep learning curve for me..

I assumed that any camera you focused on would have very good resolution available to it and any time it recorded to the hard drive it would record full resolution as well. Maybe that's not how it works.

Hi @jlppilot

You've got a complex system to setup.

I recommend getting 1-2 varifocal cameras along with an PoE switch and playing around a bit.

Look for the list from @SouthernYankee and pick a 4MP starlight+ model. if you want you can also pick up a 8MP model and compare and contrast them.

 

[jlppilot]

Hi @jlppilot

You've got a complex system to setup.

I recommend getting 1-2 varifocal cameras along with an PoE switch and playing around a bit.

Look for the list from @SouthernYankee and pick a 4MP starlight+ model. if you want you can also pick up a 8MP model and compare and contrast them.

Thanks. I have been looking for something to buy to just dip my feet in the water. Your suggestion of what to purchase to get my feet wet is much appreciated.

I realize this is a monster project and I expect to do it in phases. I just want to make sure that any step I take does not make it so that at some point in the future I have to re-do everything. I would like every component I buy to be robust enough to be useful in future situations as the project evolves.

Do you have any suggestions on a good PoE switch? I think I would really like to start with a PoE switch that would work in Building 4 @ Location A. Since I can imagine 10 cameras in that location alone at some point and since I would like the ability to add other devices there (maybe sensors or switches that I can monitor/control either over the LAN or over the internet) then I suppose I would need a switch with more than 10 ports?

I am WEAK on networking. I am studying it but it's going to be a while before I get to where I can make really good decisions.

 

[Tom S]

I think if it was me I would want to think about putting a camera(s) at a point where your bad guy is most likely going to be directed into a well lit area at fairly close range if you want to get good images that really do provide facial ID. Not the kind of ID to where you can recognize your friend, but the kind that let you take a image from your system and pick out the same person at the store. If someone was coming in the yard looking for thing to steal or buildings to break into where are they going to be drawn to or funneled through. Can you put them in good lighting where their face will be lit up and exposed to the camera as they approach those areas? Is there a similar point where you could do the same with a vehicle that they might arrive in.

 

[mat200]

Thanks. I have been looking for something to buy to just dip my feet in the water. Your suggestion of what to purchase to get my feet wet is much appreciated.

I realize this is a monster project and I expect to do it in phases. I just want to make sure that any step I take does not make it so that at some point in the future I have to re-do everything. I would like every component I buy to be robust enough to be useful in future situations as the project evolves.

Do you have any suggestions on a good PoE switch? I think I would really like to start with a PoE switch that would work in Building 4 @ Location A. Since I can imagine 10 cameras in that location alone at some point and since I would like the ability to add other devices there (maybe sensors or switches that I can monitor/control either over the LAN or over the internet) then I suppose I would need a switch with more than 10 ports?

I am WEAK on networking. I am studying it but it's going to be a while before I get to where I can make really good decisions.

Hi @jlppilot

FYI - we have a list of PoE switches, and a thread on this topic. Just start with a good smaller one for now for testing..
( hmm... check the wiki section or cliff notes for the link, I'm short on time right now to search for the link .. )

 

[jlppilot]

I think if it was me I would want to think about putting a camera(s) at a point where your bad guy is most likely going to be directed into a well lit area at fairly close range if you want to get good images that really do provide facial ID. Not the kind of ID to where you can recognize your friend, but the kind that let you take a image from your system and pick out the same person at the store. If someone was coming in the yard looking for thing to steal or buildings to break into where are they going to be drawn to or funneled through. Can you put them in good lighting where their face will be lit up and exposed to the camera as they approach those areas? Is there a similar point where you could do the same with a vehicle that they might arrive in.

Good plan. Footprints from the recent robbery indicate they took a totally different route than I would have anticipated. Really frustrating having to try and think like a thief.

 

[jlppilot]

Hi @jlppilot

FYI - we have a list of PoE switches, and a thread on this topic. Just start with a good smaller one for now for testing..
( hmm... check the wiki section or cliff notes for the link, I'm short on time right now to search for the link .. )

Thanks. I am searching through this site for the resources you and others have mentioned now.

 

[ip_poe]

@jlppilot Might want to check out: IPVM Camera Calculator V3

Click "Change Location" and locate your property. It let's you only do up to 4 cameras for free, but I just move the same ones around to get an idea for each area of interest. You can get a good idea of what a face or vehicle might look like at a given distance, resolution, lens width, PPM, etc. For license plates (Especially at night), you'll need an LPR (License plate reader) capable camera, there's a separate master thread here with those details. Ideally you'd put an LPR at points where a vehicle must travel through (a road or gate that leads to the main entry to the property and that traffic must pass through).

 

[looney2ns]

This camera "if it will give the PPF you need" is a good starting point and does well at night.

Review-OEM IPC-T3241-ZAS 2mp AI Lite series Varifocal

IPC-T3241-ZAS from EMPIRETECANDY 2MP WDR Turret Varifocal AI Network Camera Review unit courtesy of Andy (@EMPIRETECANDY) in exchange for a fair and honest review. Purchase in the IPCT Store Andy's ipcamtalk vendor forum: link Andy's AliExpress store: link Andy's Amazon store: link Price...

ipcamtalk.com