I have an urgent enquiry for border secutiry force, asking for identification of human beings at about 4km. Taking ppm in to play is there any CCTV camera that will do the trick. What about Avigilon professional cameras?
Subscriber Discussion
Identification At 4KM For National Security
JH
John Honovich
Nov 15, 2014When you say 'identification of human beings', what type of image detail do you desire? 1, 2, 3, 4:
To give you a rough sense of what you are looking at / for. At 4000m, with a 1080p camera, and a huge 1000mm lens, best case scenario you will get this:
I am pretty sure it will be worse than this because of lensing / atmospheric issues at such super long ranges plus really long lenses tend to have high f stops and/or are fantastically expensive.
For more color, take a look at: View A Face 3000 Meters Away.
In general, you should be looking at long range military focused products. For example, Vumii.
TG
Tedor Gligorich
Nov 16, 2014Side noting: that you get 99.1 ppm even when make lens 790mm all way up 1100mm.
Never did I catch this before, sorry, because used to be that "cap" would kick in at a Rukminimum of 500mm. AOV probably need more decimal precisioning.
Marc Pichaud
Nov 15, 2014You probably have to prove it's human being crosing the border, and get a Johnson rules footprint not to identify them. As John says it's a differnent story especially in darkness
What soldier use in Afghanistan, to identify Allies or ennemies and if your carry weapons (and sometimes shoot friends, oups..) are thermal cooled systems with very long focal. See on Flir side (35 to 60 K$ per camera)
Not sure you will be able to distinguish Mexicaners from US citizen if they don't wear sombreros..:-)
Ray Bernard
Nov 16, 2014Roshan, you may be able to work with Apex Corporation to design something that will be helpful to you.
I have very successfully field-tested their Q83 Hydra Pro PTZ product. an analytics-based multi-camera PTZ, and I have specified it for two 2015 wide-area video surveillance projects. You can download a brochure from the product page.
Multiple cameas are contained in a single PTZ dome housing (PTZ camera, up to four overview cameras). The PTZ camera is automatically controlled by analytics running in a separate appliance.
I originally learned of this analytics-based product when I came across its deployment with analog cameras at Teberboro general aviation airport in New Jersey near Manhattan. For low-traffic high-quality close-up images of people and/or vehicles, it is an outstanding application.
Apex purchased the analytics software from its original maker, and upgraded the technology to work with analog or digital video cameras. So you can have 4 HDTV overview cameras (totaling about 8 MP for the overview). The standard product unit can be configured for overview camera total field of view ranging from coverage of any field of view ranging from 180° to 360°, but I needed a more narrow long range overview area field of view, and they worked with me on camera selection for that.
In our field tests we were able to zoom in on people, vehicles, and bicyclists at 800 ft and get the target person or vehicle to fill the 3-plus megapixel PTZ image. I know you have a range requirement that's about 16 times my 800ft requirement.
However, and amazingly, because you need only 2 pixels per foot (about .6 pixels per meter) at the far extent of the target field of view, I'm suspecting that Apex may be able to give you, for example using an 8mm lens with the HDTV overview cameras, 40° coverage per overview camera--giving you .66 pixels per meter at 4K meters.
Typically, the larger the potential order, the greater the motivation of the vendor to accomodate your needs.
They offer live demo videos from 9AM to 4PM on their website.
Apex has a mobile field demonstration van that they have taken all around the U.S. for field testing purposes. That's how I got my field tests done.
At the time I did my testing, it was $5K for the standard Q83 Hydra Pro PTZ product that included the PTZ camera, four fixed overview cameas, and the analytics unit. Each Q83 unit gets its own analytics box.
The technology uses PTZ presets to control the PTZ camera. The system does a very good job of jumping back and forth between two objects in different PTZ preset zones. More than two objects can be tracked but you get of course less time per target object, due to PTZ camera movement. For my applications, this worked out just fine. If you need to track more than one object (or group of objects within a target zone), you may need some custom tuning of the parameters fed to the analytics algorithms. I discussed that with Apex a little, but never walked down that path.
You can record the streams from all of the cameras in your VMS system.
Lessons learned from field testing:
- The camera configuration requirements can be significant:
- In the overview cameras, you want to mask trees and shrubbery that may be subject to wind motion.
- The technology uses PTZ presets to control the PTZ. I found that high resolution images of the intended camera locations are useful for plotting the PTZ preset target areas.
- To achieve consistent PTZ image quality, you have to consider the image pixel density at various distances from the camera. So you may have more preset zones further out than you do closer to the camera location.
- I found that my desktop calculations/estimations were good but not as good as the results from refining them working with the actual product on-site to figure all this out and verify the configuration.
- Lean on Apex for all the help they are willing to give. They have lots of relevant experience. Some of the work is non-intuitive.
- Don't worry about making mistakes in your own thinking or calculations. Apex support will cover you in this regard. That also means don't take your thinking too far without consulting with them. I gave myself some needless headaches.
- Sometimes it is easier to reach Kaytie than Michael, due to his travel and work schedule.
I worked with their CEO Michael Mojaver for getting the engineering work done and for field testing execution, and Account Executive Kaytie Storm for logistics for field testing.
Michael Mojaver
President/CEO
Apex Automation, Corp.
2890 S. Santa Fe Ave
Suite 101
San Marcos, CA 92069
(760) 305-7289 direct
(760) 295-2227 FAX
Kaytie Storm
Account Executive
(760) 571-9101 direct
(888) 939-APEX x220
JH
John Honovich
Nov 16, 2014Ray, I am confused about how this is a fit here.
Specifically:
"However, and amazingly, because you need only 2 pixels per foot (about .6 pixels per meter) at the far extent of the target field of view, I'm suspecting that Apex may be able to give you, for example using an 8mm lens with the HDTV overview cameras, 40° coverage per overview camera--giving you .66 pixels per meter at 4K meters."
Where are you getting that he only needs 2ppf? Are you saying this Apex company is detecting objects with just 2ppf? I don't think even a human could reliably detect a person with just 2ppf in perfect lighting conditions.
Also, an 8mm / 40° lens, even with 5MP, at 4000m delivers just .3ppf, not 2ppf.
Ray Bernard
Nov 16, 2014John, I was using the lens calculator in Metrics mode, which gave me the data I reported originally for the 4000m camera reach.
When I put it in English mode and used an online metrics conversion tool, I got 13,123.36 ft for 4000 meters. When I put the 13,123 ft in the lens calculator using a 1080p camera and 40° field of view with 8mm lens, I got 0.2 PPF similar to what you said.
I don't understand why the metrics version didn't work correctly, unless I am somehow misreading the data.
At 6,500 ft (about 2000m) I got 2.01 PPF with a Theia SL940 lens set at 40mm, which gives an 8.4° field of view. That's over a mile away. The horizontal field of view is 955 ft (291 meters) per the lens calculator. So per my experience there is a good chance this would work at that distance, although that's a much more narrow field of view than typical Apex applications, so this unusual applicaton would have to be field-tested.
If Roshan contacts Apex there may be a potential solution. They are a bit conservative about doing something outside the parameters they have used in previous deployments. That's something I like about them.
If you think my postings are still in appropriate, just forward the comments directly to Roshan and delete the two comments. That would be fine with me. My primary purpose is to be helpful to Roshan, not mecessarily to promote Apex (even though I have had good experience with the product and company).
JH
John Honovich
Nov 16, 2014"I don't understand why the metrics version didn't work correctly, unless I am somehow misreading the data."
If you get .66ppm, then it's going to be ~.2ppf
Because if you have, e.g., 3.28ppm, you will only have 1ppf because a meter is 3.28 longer than a foot. I think you multiplied by 3.28, rather than divided.
In any event, I think we agree now that, 4m away with a 40° AoV is way too wide angle to detect anything.
As for Apex, my main issue is that, as it is described, I do not see how it could possibly deliver what Rohan is asking for, "identification of human beings at about 4km"
Ray Bernard
Nov 16, 2014John, I am glad you straightened me out on the feet vs. meters.
If the requirement is to "detect motion of a person-sized object or small group of people" and track them, it may work if the environmental conditions are suitable and the analytics will work with very narrow fields of view.
On the other hand, if the target environment has sagebrush being blown around, for example, it won't work because you will get too many false positives. If there is airborne sand much of the time due to wind conditions, it won't work. Air quality becomes a consideration when dealing with long-range lenses and great distances. Another distance issue is camera stability, because a tiny movement of a pole-mounted camera can move the field of view by many feet on the far end.
If Roshan can get more information about the environmental conditions and exactly what is trying to be detected, that would be helpful. I had assumed, but didn't state my assumption, that people will be involved in the identification aspect. I thought the Apex technology had a chance of working because it uses the PTZ to zoom in on the detected moving object, and an individual can very easily tell the difference between rolling sagebrush, a donkey, a person, a group of people or a vehicle when the target image fills the screen. You can't recognize much at very far distances with the overview cameras. But the PTZ image is a different story.
Enough said on my part. If it were this easy a challenge to solve, the solution would already be in place given the border security needs and the potentioal provider revenue involved.
Ray Bernard
Nov 17, 2014Roshan,
I heard back from Michael Mojaver of Apex Corporation, who told me that Apex has a customer that OEMs some technology from Apex, and produces cameras and surveillance platforms designed for border surveillance applications as well as other long range applications.
VIT Security Group is the company and Longview 2 is the product.
JH
John Honovich
Nov 17, 2014Ray, I think the Longview 2 might have potential here, in that it is, at least, a long enough lens to target down that far away.
That said, I continue to believe that surveillance at 4KM is either going to need its own light source built in (like Vumii) or non light (like thermal).
In particular VIT's claim here is questionable:
"The VIT 53 series of ultra-low-light (0.00003 lux) cameras with ‘night mode’ produce unbeatable evidential images in day and night conditions. Developed during the ‘Cold War’ they operate at a sensitivity of over 150 times greater than traditional cameras."
Besides the fact that lux ratings like that are marketing nonsense in general (0.00003 lux might as well be a black hole), they are clearly not factoring in the impact of their super long lenses, which is going to have a high f number, meaning an 80 / 90 / 99% reduction in light passed.
New discussion
Ask questions and get answers to your physical security questions from IPVM team members and fellow subscribers.
Newest discussions