Noting that IndigoVision is selling an apparently NOX-20 derived camera, obviously with integration to their VMS. Also ONVIF compatible.
Selling Super High Resolution Cameras
Is bigger better?
In the past decade, the surveillance industry has moved from 0.3MP to 1MP to 2MP to 3MP to 5MP to 4K (8.3MP), etc.
But there are even higher resolution cameras above that. This report covers selling those super high resolution cameras, those that deliver beyond 12MP.
Super high resolution cameras share many of the same pros and cons of selling 4K cameras, with both the pros and cons being even more so when applied to higher resolutions.
Additional Pros Of Super High Resolution:
- Demonstrate "product leadership"
- Eliminate competitors
Additional Cons of Super High Resolution:
- Higher proportion "wasted" pixels
- Price to pixel ratio
- Limited CODECs
- Proprietary or Limited VMS Options
Good article for all us high res fanatics!
One thing I'm not following
Higher Proportion Of Wasted Pixels, Increased Storage Costs
Although I see that more pixels are possibly wasted in SHR cameras, why would the proportion be different?
Let's use a large parking lot as an example. In the image below we have some piece of the world, where there is a parking lot in the front of the image, and another parking lot sort of in the middle right.
For the purposes of this application we can assume that a single 30MP camera with a 16:9 aspect ratio gives us sufficient detail on the further part of the image, and more than sufficient detail on the near part. This 30MP camera is represented by the red rectangle, which you can see is also capturing a lot of skyline, the building on the right, other parts of the road, etc.
The 4 blue squares would be 4K cameras, setup to cover just the parts of the scene that we actually care about. In the near field the PPF would be roughly similar, and like the 30MP camera more than sufficient. In the further parking lot we could actually get even better PPF than what we achieved with the 30MP camera (due to the ability to crop the frame to just that area with the right lens).
Also, given that the 30MP camera delivered sufficient detail in the far area originally, we might be able to get by with 2 1080p cameras to cover that area, which would reduce system price and storage costs even further.
Because we can adjust multiple cameras to cover the scene with more flexibility than with a single imager, we can reduce the percentage of wasted pixels as well.
What this comes down to is that many times what we want in security, and artificially create, is something like a 32:9 or 64:9 aspect ratio. This is partly why multi-sensor cameras are more popular than super high resolution cameras, they give you more ability to spread a fixed amount of pixels over a given scene optimally.
Yes, agreed when you are replacing multiple lower resolution cameras with a single higher resolution camera. In the case when you are upgrading to get additional details, i.e. not changing the scene, I would think it would be the same proportion, although a greater absolute number of pixels wasted by SHR cameras.
One advantage to the single HD camera is you are covering the areas around the parking lots so if say a car is stolen you will be able to see which direction it went and maybe them casing the area which you would miss with lower res tighter FOV cameras.
Also, I always find it easier to review video on one camera instead of switching between cameras
Has this market continued to open over the last year with more offerings and more compatibility with vms.
i have an 2 car wide alley that I need to cover from a building at the end of it, distance is 489 feet from camera location to end of alley. Requirements state I need at least 40ppf on target from the begging to the furthest point (489ft).
By my rough calculations I will need at least a 25MP camera.
Any idea or suggestions?
Ud1 yes but I need to maintain coverage from 115ft to 489 feet. I could get either with a lens change but not both necessarily.
I have the added challenge that it is an historic facility thus visual impact has to be kept low, I have thought of some creative camouflage but would like to stick with a box style camera verse a dome to minimize the footprint facing the street.
Depft of field is going to be a problem if you need the whole image in focus from 115ft to 489ft from the camera. What VMS are you using?
Perhaps a camera with continuous autofocus or a multi-imager then? (Though I see the footprint/dome issue, recently raised)
Kind of thinking along the same line since it becoming very difficult to locate a single all-in one box camera at the needed resolution. I am going to contact Bosh in the morning to look at the MIC-9502-Z30. This camera does not appear to have been reviewed nor is in the calculator yet. Looks promising with a number of useful features and what appears to be a style I could build a concealment around to blend into the architecture.
Since thread has picked up activity a year later, I thought I'd mention that Entropix will be debuting our neural network based Extreme Resolution on Demand technology in Washington DC next week at NVIDIA's GTC DC show which is focused on their Metropolis (AI Smart City) initiative.
It addresses most of the pains associated with resolutions of 4K and higher through a different dual-sensor approach with GPU's utilized in the pipeline. Let me know if you want to go and I can get you a discount code for your ticket.
What is the projected availability and price of a surveillance camera based on this technology? Also, is the plan still to make the resolution enhancement piece subscription/on-demand based, or will it be more fixed price in surveillance applications?
Projected availability for purchase of cameras at volume is Q2 of next year. (Releasing at GTC and ISC next year). A dual 720p (4K on demand) and a dual 1080p (6K on demand will be available then). A dual QHD (8K on demand) will follow within six months of the release of those two. Additionally there will be a custom camera utilizing dual 12MP capture at low frame rate (108MP on demand) as well as a dual 720p body-worn camera right around the same time.
Pricing won't be shared yet for cameras, and each variation has it's own licensing model per its application. But all models will still revolve around some licensing or on-demand subscription model in terms of how we charge for it. However in some cases, our customer is the camera manufacturer, so how they charge their customers is their business. Entropix however is a software company so our business model revolves around licensing, not hardware sales.
Happy to provide more details offline. Right now our upcoming technology debut in DC is geared towards higher level video applications (the nature of NVIDIA's show) and we're just going to be showing the technology integrated into NVIDIA's new DeepStream SDK model as part of their Metropolis Initiative. The DC show is devoted to this initiative.
...and a dual 1080p (6K on demand will be available then).
So the idea is that I pay for a camera in which I normally view just one of the sensors, but record two, and then can pay more to triple the number of pixels in both recorded streams to view at high res?
You pay for one more sensor and increase your bandwidth by 65% (monochrome stream encodes more efficiently). This gives you an extreme dual sensor day/night camera with no IR cut filter and true monochrome encoding. Examples of such cameras in the market already are Mobotix MX6, Arecont 3236DN (eg not new hardware concepts).
(an aside...) This hardware concept has now also been mainstreamed by nearly every dual sensor smartphone manufactured with Snapdragon 835 processor including Moto Z Force, Android Essential, and a half dozen others plus the Huawei P10 and Mate 9 on HiSilicon's Kirin chips.
With a dual sensor camera configured to our open reference, plus our GPU software, you get 9x the number of pixels (3x vertical/3x horizontal). Our software is designed from the ground up for a pixel consumer base that is not, in fact, human. Machine vision has no interest in the blue sky, the empty asphalt, the green grass and trees, or the static blank walls. Much like human security video specialists, the only pixels the computer vision cares about are 'pixels of interest'.
Our model allows resolutions up to and well beyond 100MP on demand in order to extract high order resolution on target pixels (as defined by a computer vision) while handling the irrelevant pixels in hyper compressed original resolution. Like a smart codec on steroids.
A happy byproduct of our technology is that if you actually want to achieve extreme resolution video at frame rate (such as 8K at 30fps) for every single pixel, it's simply a matter of throwing more GPU power at it or waiting longer (than real-time) for the output.
Here's a basic scale:
Dual 720p = 4K resolution on demand. Extremely low cost camera. Large pixel format for much greater low light sensitivity (which is augmented by the true monochrome capture and encoding without cut filter). Total bandwidth and storage required for camera is 1.65x the normal such for a single 720p. Delivers extremely high quality, low noise 4K on demand.
Dual QHD (3.7MP) gives you 8K (33MP) Extreme UHD resolution with less bandwidth, storage, hardware cost, power consumption and heat generation than a standard 4K camera boasts.
Dual 4K (8MP) gives you 12K or 72MP resolution at full frame rate. As of yet, there's not even a decoder or monitor that can display that resolution properly so it's almost entirely a zoom in and render at one of the lower resolutions only proposition for now at that extreme.
The dual 12MP sensors in all the smartphones mentioned above will soon be able to provide anyone 108MP still shots on demand with a simple cloud app.
Hope that explains it a bit more.
With a dual sensor camera configured to our open reference, plus our GPU software, you get 9x the number of pixels (3x vertical/3x horizontal).
Yes, but isn't that really 2 (one from each sensor) pixels to 9?
Which I think is fair to count, since a similarly priced single sensor camera could have at least double the resolution.
But think of the storage cost and better low light ability. Even if it is 2x720P sensors, the storage would be 1/4 of 4K. Not only that, the low light capabilities would be far superior.