Avigilon 29MP/16MP Cameras Vs Dallmeier Panomera

Hi,

I have read here about the Dallmeier panomera and tried to see it in action but in most clips on youtube and spec sheets the data is very vague and difficult to understand what can be achived when using the panomera system. I saw a lot of people talking about the system and not many using it. The only clips that do show the use of the system are not from Dallmeier and show avarge quality.

I can't really understand what is covered by a single unit.

I'm trying to compare this to the Avigilon 29MP and 16MP cameras. With Avigilon I can cover about 30-40 meters width and 25 meters height and get facial regocnition. What is the area size that a 51MP Panomera unit can cover and get a good facial recognition. Also what is the FPS of the 51MP when compared to the 2 FPS of the 29MP and 10 FPS of the 16MP?

Last is the 50,000$ "street price" of the panomera unit is real or exagurated?

Is there a demo clip with a player that one can get a feel to what you get when you buy this unit?


Itamar, Dallmeier does have a few real demos here, but they only show the 4 imager model.

One reason that Dallmeier is vague is that performance / coverage depends on the number of imagers and focal lengths choosen.

According to Dallmeier's promotional booklet, which has the most details I have seen, "models with 4, 7, 8, 14, 15 and 17 modules are available." And those models can be oriented in different ways - landscape, portrait, etc.

These are a few examples of sizes / form factors:

Also, the lens selection impacts coverage area / pixel density, etc.

Here is an image that displays 3 different focal lengths used in a 4 image model:

Notice that there is 1 wide / full imager, one in the 'middle' of the image' (#4) with a longer lens, and two up 'top' at the far field of view (#2 and #3) that have the longest lenses.

To your specific questions:

"What is the area size that a 51MP Panomera unit can cover and get a good facial recognition."

Answer: Depends on what model you choose, what area and what focal lens lengths. I am pretty sure you will have to talk to Dallmeier to get clarity on this.

"Also what is the FPS of the 51MP when compared to the 2 FPS of the 29MP and 10 FPS of the 16MP?"

Answer: Dallmeier says up to 30fps for all imagers, because they are using multiple lower / regular 3MP imagers instead of one 'giant' imager.

"Last is the 50,000$ "street price" of the panomera unit is real or exagurated?"

Answer: Depends on the configuration choosen, number of modules, etc. but I would expect it to be in the tens of thousands USD, given the number of imagers involved, the custom built, Dallmeier's typical high pricing, etc.

Btw, both options are proprietary so enjoy the lock in :)

hi - just a comment: Panomera is onvif-conformant device (see onvif site, product == panomera streaming server). Or, to be precise, there is an onvif adapter for it, that makes multiple conventional 1-5 megapixel streams (from different ptz pozitions) - something that scallop is doing, but in quite another pixel budget????. So, it is not pure proprietary.

Alexander, thanks for sharing, I stand corrected.

As you mention, the Panomera camera itself is not ONVIF, it is the Panomera Streaming Server (see ONVIF listing). It is not clear to me whether the 'Streaming Server' is built into the camera itself or is an additional device/unit.

From a Dallmeier document, here is a description:

"A Panomera Streaming Server provides a single Full HD stream that comprises the various Sub-Channels (streams) of a Panomera multifocal sensor system.

The current version of the SMAVIA Recording Server software [Dallmeier's own] now supports lossless zooming in this stream with the external software SMAVIA Viewing Client (SVC). And as a digital PTZ function, the zoomed area can be placed anywhere."

I do wonder how it all integrates with third parties, especially the switching to the longer focal length imagers.

server is an additional device.

and it is onvif device. You can think about it as a multi megapixel camera with multiple digital zoom units, that can magnify 80x and still deliver D1 stream with pixel to pixel precision.

so, from vms perspective this is not a big difference from multi-stream adapter from analog ptz cameras, that you described about a week ago - thats onvif.

and on this level there are not word about lenses - they are not varifocal.

Thank you John.

This cleared some things for me.

From your description (which is much better than Dallmeier) I understand that the 51MP cameras unit is probably composed of 17 X 3MP cameras. In case you would like to cover a large crowd in the ditance like in a soccer stadium you'de have to use all the cameras with long range lenses like 100mm or more.

This would probably work well at day time when there is a lot of light and poorly at night since the long range will cases very little light to go into the little lens -no? At least when compared to the the Avigilon solution that uses a full frame SLR lens.

Do I make any sense or are my thoughts far fetched?

Itamar,

I am not sure how long of a lens will be needed for a soccer stadium. It depends how narrow of any area each 'module' covers and how far away the camera unit is.

In general, longer the lens, greater the risk of higher f number, ergo worse low light performance.

On the other hand, each Dallmeier camera is lower resolution, which means greater probability of improved low light sensitivity (compared to a single super high resolution). Also, many of those SLR lens used with Avigilon have fairly high f numbers (or even more astronomical prices).

Net / net, there are too many variables cutting both ways for me to make an armchair guess about which one is better in low light.

Btw, Avigilon rates their new 16MP H.264 CMOS model at 20x the lux rating of their older 29MP CCD model (spec comparison), so appearantely Avigilon thinks their new 16MP low light performance is a lot better than their legacy 29MP one. As such, if low light is important for you and if you go with Avigilon, that difference should be strongly evaluated.

Here is a Panomera S4 (4 imager) data sheet with more details. A few things standout:

  • The 'widest' FoV is just 22° horizontal.
  • Claiming a max frame of just 12.5fps (contrary to the 30fps claim in marketing docs)
  • They have created a weird 'effective resolution' metric, claiming up to 40 'effective' MPs, even though these are (4) 3MP or 4MP imagers.

can't say for sure about this model - never seen it in action, but i know that they are using common platform with 30 fps sensors, so maybe this is cheap device with (relatively) weak cpu.

as for effective mp - the idear is clear: They can distribute pixels equally per feet on different distances, and refer to conventional cameras without this feature. So, to have the same dencity on far distances, conventional cameras need more mp.

Where Dallmeier is claiming the most 'effective' number of pixels is in telephoto / narrow scenes. For example, their 7 imager model covering a wide 95° only claims 45 'effective' MPs, while their 8 imager model covering a narrow 12° claims a whopping 146 'effective' MPs.

It appears they are using a number of longer lens imagers and counting them as X number the pixels of a regular camera.

"So, to have the same dencity on far distances, conventional cameras need more mp."

Or you can just use / add more longer lens cameras focused on key targets farther from the camera.

My main concern is that introducing a term 'effective pixels' is going to create confusion because what they are doing is not magically increasing pixels but adding additional imager / longer lens for specific subsections of the scene.

exactly!

my understanding of their point is that they provide uniform way to access these complex field of view.

with conventional cameras you either need to use multiple cameras (loosing ability to smoothly move from one place to another, required for example for tracking people), or to use multi-megapixel camera (wasting pixels attached to near field) with digital zoom on vms side. Their marketing megapixels are megapixels of these imaginable camera. Thanks for refering to 7/8 sensors camera - their booklet now contains both resolutions (cumulative sensor resolution and effective resolution).

from my point of view their major weakness is another side of their strength - singe point mount. This can be partially mitigated in huge installations by covering zone domino-like.

It appears they are using a number of longer lens imagers and counting them as X number the pixels of a regular camera.

Replace one 29MP camera with seven 4MP ones!

Running the Are-vigilon pitch in reverse!

nice joke!

what they are claiming is that you can use 4 (four), not seven. And smoothly ptz-navigate them all. And with full or half fps. This is one camera to the outside world, not 4 or 7 (like scallop, not like multi-imagers from a-t/a-n).

Ok, even better, I didn't see the 4 imager model at first. :)

I think I get the math now, though I agree with John that the term 'effective megapixels' is a tad misleading. One reason: it's debatable to what degree you are really

wasting pixels attached to near field

with a single imager camera. You say wasting (I assume) because the nearfield PPF are higher than the minimum PPF for the whole FOV. But more nearfield detail is still better than none. And some real-world requirements dovetail nicely with the concept of increasing PPF with decreasing distance, since closer objects are more of a threat. Object detection at 1000M, Person detected at 300M, Facial at 100M.

That is not to say I don't see a large benefit to their 'pixel reallocation' strategy, I do, it just seems too scene dependent to assign a simple metric like EMP.

One other question, when panning 'seamlessly' across stitched views at the highest resolutions, is there any 'google maps' patchwork effect (at the imager transitions), due to the different lens distortions or resolution differences?

as for your last question: Can't say for sure for 4/7/8 models with near/far ratio about 5-8. But for their main 17 camera device (ratio about 3) that i played at se'12, there were no such effects. They were saying that images are captured at really the same time, and stiched quite precise. Maybe not with pixel to pixel precision, but quite near. It was possible to find stiching line, but mainly due to a bit different white balance/color settings on nearby cameras (a side effect of wdr, i guess).

please take a look into airport demo provided by topic starter: There you can see overview stream (bottom) and magnified stream, and distances at left and right are different.

Here's the airport demo again:

You can clearly see at least 4 separate imagers as the operator scans from right to left.

I don't see how they can have that many more 'effective' MP here because of the super wide angle FoV (as they admit in their own documentation/calculation).

In such an application, an Arecont SurrondVideo camera would be quite similar.

I agree John,

I don't understand that too.

I can only imagine what they mean is that if you used conventional cameras you would get "wated" pixels" becuase you would have to use a lot of similar cameras at different FOV to cover this area with overlapping space.

But it seems to me that this is in fact what they are doing but very precisely and so not many pixels (if any) go to waste.

Sorry - let me repeat, they are introducing this metric only for "corridor mode" camera models. And they are comparing single panomera camera with imaginable single multimegapixel camera that has the same pixel density at far distance. This camera will have much bigger pixel dencity on short distance, while panomera will have more or less the same. As "undisclosed a" mentioned above, there are situations when short distances are more important - these cases are not fitting to panomera concept. As i am not a person of surveillance practice - i have no opinion aboth what case is more typical, and, btw, i am interesting in yours.

oops. Airport look like an example of isopixel camera set. Strange, as right corner should be captured by mre sensors than left. Anywhay, this is an example of how pan/zoom works.

i expect than if you compare it with arecont as two closed solutions, you will find not a big difference (i am not familliar with arecont in details, but i expect they have full rate and nice stitching in their vms and few partner vms). Onvif add-on make a small difference.

i'just followed your links to their recent demos: It look like "river elbe" is most relevant for zooming. And please note right-top corner: Look like they left it without camera.

and all that mp counting magic applicable only to sets with different fl: So airport is not relevant here, but river is ok. Let me repeat - they are refering to mp of a single conventional camera, mounted to the same place and capable to look at arrows on clock tower. This is also not nice illustration, as their strength when you need both far and near, that is not the case for river demo.

for me the outcome from that discussion is that it is extreemly hard to explain by words what you feel when you really play with it. Maybe that is the reason why they (dallmeier) are so closed in showing this baby (in europe they are showing the system in their track-showroom, under supervision and without any photo/video recording - at least for people from streets).

what really caught my eye (i am an engineer) is the concept that you do not need more megapixels than your monitor have. But you need an ability to magnify, keeping this (presumably high) resolution. And you need an ability to pan/tilt in as big picture as possible, with any selected magnification. You can have it with one multi-megapixel single-imager camera, or with multiple well-stiched cameras. First option waisting megapixel dencity for near field (in concequence waisting bandwidth and disk storage), plus it is hard to have high bitrate, plus it is waisting vms resoursed as you do not need all that megapixels - you need 2mp as your monitor have.

second option is lacking smooth object tracking, and/or require close integration with vms.

what dallmeier is promicing is to merge best sides of both. Not clear how far they are on that way.

Here is the 7/8 module units, which support wide FoVs and claim 30fps max.

Thanks for pointing these new devices!
And now it look like i got your point aboul lenses. Yes, for long distances lux rating should be worse than for near. But that is not something you have not with normal camera, with the exception that far objects will be not only dark but also out of focus.
Btw, as i have seen their basic design (security essen 2012, really impressive and eyes capturing, managed to see and play: Magnifying to far objects and still have no pixellization on 4k display), they have different lenses for different sensors within the same device. I am not an expert here, but maybe it is possible to compensate long distanses with power optics.
As for niche, they are now mentioning mass transit platforms - an ultimate "corridor mode".

The reason I say it is a niche is simply the much higher price point of this vs single imager cameras. There are $1,000 4K cameras on the market now. I think most will be hard pressed to pay 5x, 10x, 20x more for the Dallmeier system.

On the other hand, I do think it's a smart move for Dallmeier as even if it a niche, they have a differentiated product to offer for it.

btw, maybe you can remind me about US company that is selling their camera to government and giving for rent to others (accompanied by an aircraft????). They are using 16 photo cameras in the same manner as dallmeier, trading fps for hi-quality long-distanse lens.

Like this, (with another row)?

yes, right - this civil version of now disclosed argus.

thank you a lot!

Thanks guys for the great dicscussion about this.

I too do not like the effective pixels term. Its confusing and misleading, reminding me the term of dynamic contrast in LCD.

Anyway the bottom line is WYSIWYG and I can find demos of the Avigilon 29MP camera but absolutely 0 from Dallmeier. How come there are only youtube clips and no exported clip with Dallmeier player that you can play with and see for yourself?

I really want to see the 30 FPS and what happens when I zoom to the far side and what happens when I zoom in to a near side. The focus behaviour and low light capabilities.

Or in other words:

i totally agree that for any non-conventional thing you need to try before spending money.

maybe they have panomera available online via onvif? You can try to ask them.

otherwise you can examine airport demo from your starting post, or meet them in an exibition nearby your location.

Btw, as a point of reference for the US market, Dallmeier said one of their US partners would be showing it at ASIS 2014. Ethan went to that booth and they had no Panomera units and the people there at the time had no details on it.

Of course, since Dallmeier's home market is Germany, I would expect more info / accessibility in Europe.

If IPVM did only one test next year, here's my vote:

Maybe call local Tri-Ed branch to see if they have a 'open box' loaner in the backroom?

is it a real photo? Wow! Pretty penny in one place (left unguarded).

btw, John: What do you think about the concept that resolution of monitor matters, and resolution of camera don't (until it is, by some magic, capable to deliver enougth data, and all area is covered with enougth pixel per feet density)? So the way we are really doing this are for historical reasons (a bit simular to alien's view on ip video from dvb-t camera discussion).

I too am getting the feeling that the Panomera system is best suited for corridor mode.

When you benefit from small number of cameras devided with different focal lengths and stiched together.

On the other hand I do not see any benefit (except auto stiching) that the Panomera system have when looking at one single far away object like in this example:

I really don't see how looking at a soccer crowd that is 70-100 meters from the camera will benefit when using this system.

The road show is another matter. While the unit and truck look amazing again I don't see any footage of actual video.

One word to IPVM (John) if you can get your hands on a Panomera system this would be a really really nice review that as you can see it will draw a lot of attention.

it's all about pixel on target that you expected, both Avigilon 29MP/16MP Cameras Vs Dallmeier Panomera are different purpose, hence details vs. situation awareness. I will go fisheye with dewarping as alternative cheaper solution.

No doubt fisheye with dewarping is cheaper (much cheaper) but the big drawback is that fisheye cameras will always deliver much less pixels on target at any distance than these two cameras being discussed.

I am not saying fisheye is bad and fisheyes are obviously far more used than both Avigilon PRO and Dallmeier Panomeras combined but fisheyes are really for close up monitoring / overview awareness, not long range pixels on targets.

Newest Discussions

Posts Latest
4
less than a minute by Jonathan de Chateau
3
less than a minute by Ross Samek
13
about 2 hours by Undisclosed End User #5
69
about 3 hours by Undisclosed #2
2
2 minutes by Shannon Davis