ARM based Image Processing system

If it exists, your customer doesn't need you.

You will likely need an ARM with external memory interface.

Try the Phillips SAA71XX video decoder chips.

What is the purpose of the prototype? (Are you trying to demonstrate specific performance at a given price point, or is this an academic project, or what?)

What are the constraints of the system, besides what you've listed below?

Why did you specify a CCD camera, and why are you talking about composite video input?

Based on what you put in your original post, the path of least cost and least resistance is to use a PC - or a single-board computer PC - as the host, and connect either a composite video capture card or digitally connected [USB] cameras. Focus your attention on where you're adding value for the customer; i.e. the software.

You really don't want to design this hardware from scratch, particularly if this is not your forte. Unless there are power or size constraints or something else you haven't told us, there is no good reason to go with a specialized system; use off-the-shelf PC components and save yourself a lot of money and trouble.

If you insist on going the complex route, look at a high-end FPGA; I'm thinking the Xilinx Virtex-4, for instance - everything you want is there except the ADCs for the video.

I guess the customer asked for both.

Or a video processing SBC.

And the memories. He's going to need lots of memories for signal processings, compressions and encryptions. FPGA is not very efficience in making memories.

Oh, sure, but by "everything" I meant "all the application-specific possibilities". I agree he will need external flash and SDRAM.

Hi, Thanks a lot for the suggestions. I guess the information I gave does no define the problem completley. Let me please add the following information:

i) I am working on an indentification project, Where I capture the facia image of a person process it, and store the template in the db. Later thi image will be used for identification.

ii) If this prototype is succesfull, we may plan to deploy it acros various locations on the campus, and a few other areas.

iii) This system has to be stationed at the entrance, and hence I canno go for an off the shelf PC solution.

iv) Why CCD camera with composite video? That is the kind of camera that could get to suit the requirements. (the camera has to be sensitive in th IR region). I don't need the whole video, I just need the image at th instant when the click occurs.

v) An FPGA solution may not be viable as the volumes don't justify th cost.

Please let me know if you need any more details. Looking forward to you comments.

go

It would be easier with a single frame camera, rather than a video camera.

Yes, the story is always the same.

Perhaps you should RF the images somewhere else.

There are IR camera modules as well.

What cost? I would choice a low end FPGA ($20 to $30) for sampling only, then process the data with a micro. I think you have a few seconds to process each frame, or even process them at a later time.

Take a look at TI DaVinci chip.

--
******************************************************************
*  KSI@home    KOI8 Net  < >  The impossible we do immediately.  *
*  Las Vegas   NV, USA   < >  Miracles require 24-hour notice.   *
******************************************************************

I would also like to add that I am based in India, and sourcing of man components is not easy.

You DEFINITELY want to use a PC for this; custom hardware is a total dead end that will add man-years to your project. I don't understand what you mean in this point:

What is driving this comment? Size? Power? Perception? A PC-compatible SBC that will do everything you need and operate off a bootable CompactFlash card fits inside the form factor of a 3.5" hard drive, and it does not look like a PC, it just looks like a small box. Search for

3.5" biscuit SBC to see what I mean. Many vendors sell them prepackaged in a housing with power supply, for example

You are really not just building a box with a camera on it. You're going to need training stations to photograph authorized users. You'll want to integrate that with ID card printing, perhaps parking passes and student administration records. All that stuff is easiest done on a regular PC. The smart thing to do is leverage your resource-hours efficiently and reuse as much code as possible amongst these different components. Plus, it will be much cheaper and faster to develop the code natively on a PC platform.

Most camera chips are sensitive in the IR region. To avoid color problems, they have an internal IR filter. This is sometimes removable. I have here on my shelf a box full of $10 USB webcams that I bought for machine vision projects. They have four IR LEDs around the lens to illuminate the subject, and they can see in complete darkness (from the visible spectrum point of view).

Try this simple experiment: Shine an IR remote control into any camera. You'll see that the glowing IR source shows up bright blue.

Relying on composite video adds a lot of analog complexity to your project. I strongly recommend you go with the USB webcam method, especially given your low resolution requirement.

Low volumes => use COTS. Use COTS. Use COTS. You are diligently reinventing the wheel from first principles simply because the wheels you see on the shelf are blue and you prefer a green one.

Sourcing of small-form-factor PC components is a hell of a lot easier than sourcing proprietary video-processing board components.

That was really a lot insightful.

Before I go on with my queries, please go through the link below:

The functionality of the system that I am making is similar to the on described above. We have about 2500+ students on the campus, staff security, whose access to some of the labs/locations needs to b controlled. We plan to have a central registering place, the said modul will be placed at various locations to control access(The access detail for a particular location will be pushed to the corresponding device a that location).

Now I have some more queries.

i) Yes, I have considered Web Cameras and also got boggled up by th complexity of using a CCD. Infact, I have played around with quite a fe cameras. However, I could not source the kind of cameras you mentione (USB webcams with IR LED's around). Can you please give me a link t vendors from where I can get this stuff shipped.

ii) Yes what you said is right, this system also needs to be integrate with a RFID reader(future requirement)or barcode card reader(existing).

iii) I may have the luxury of a few seconds while registering, but fo identification I will not be having that much time. It needs to happen i

2-3 secs (max). Else people may not use this system.!

iv) Sourcing from small - form - factor PC components shld not b difficult but how am I going to fit all these into one box finally? (wit all powersupplies, backup battery coming into picture.) Essentially I wil not be having any control on the product desing.

v) What I was thinking is if I can use a SBC like the one mentioned her

(one more that I found today) get a camera module that I can interface with this kit(USB cam will b great). It will help me in the prototype. And for my final design, I ca design my own PCB suiting the design requirements.

Look at

They have a CPU module running Linux. This has also an FPGA. You would build a small baseboard with an ADC function optimized for Composite Video. SAA71xx might do fine, and then you connect the output to the FPGA, which can store the frames in its dedicated SDRAM.

I am surprised that you think you can do it alone with the ARM9 but not with the AVR32 which has significant higher DSP performance.

--
Best Regards,
Ulf Samuelsson
This is intended to be my personal opinion which may,
or may not be shared by my employer Atmel Nordic AB

Un bel giorno ganti.r digitò:

Looks like the DaVinci processors from Texas Instruments are just right for your application:

OMAP is good and has a more mature Linux kernel (I think it's also part of the mainline now), but for the new projects I would stick with DaVinci, TI is boosting them a lot.

Certainly not, neither SPI nor I2C provide enough bandwidth.

With OMAP and DaVinci EVM (and most of the development platforms from TI and Spectrum Digital) the complete schematics are included, so you can just rip from there and live happy. :)

--
emboliaschizoide.splinder.com

"ganti.r" schreef in bericht news:7MCdnULovrOGGcXYnZ2dnUVZ snipped-for-privacy@giganews.com...

Have you had a look at these people:

Those camera's contain a vision software library, some free I/O and an ethernet interface.