Is this a framebuffer job?

Gene Heskett gheskett at wdtv.com
Thu Nov 7 12:39:06 PST 2013


On Thursday 07 November 2013 15:19:26 edgar did opine:

> Am Thu, 7 Nov 2013 13:26:14 -0500
> 
> schrieb Gene Heskett <gheskett at wdtv.com>:
> > Hello all;
> > 
> > I, and several others are attempting to make use of camera vision in
> > a machine shop environment, to auto position a lathe or milling
> > machine for instance.
> > 
> > My current camera is the highest priced of the logitech webcams, with
> > a nominally 5 megapixel format.
> > 
> > But for our purposes, the field of view at 2540x19xx is at least 75
> > degrees.  We need maybe 5 degrees because all we are interested in is
> > the pixel under the center pixel of the crosshairs.  This, when
> > inspecting the workpiece from 50mm, should represent accuracy's in
> > the thousandth of an inch category, and would generally be quite
> > useful to us.
> > 
> > So, is it possible to setup a framebuffer with 2 circular pointers,
> > the input pointer writing the 5 megapixel image as it comes in, and
> > an output pointer that only reads out the central 100x100 pixels of
> > the image, maybe even 240x240 but for our uses its overkill.
> > 
> > The problem as it exists now is the processing time for the video
> > image to get thru camview-emc and actually be presented on our
> > computer screens, is on the order of 3 to 5 seconds when the whole
> > signal chain has to deal with the 2540x19xx format of a decent
> > webcam.  If we could throw away all but the mathematical central area
> > of the image that the rest of the video chain had to process using
> > only 1 core of a 1.4Ghz atom processor, it seems we could save many
> > valuable seconds of image processing time.
> > 
> > The V4L list didn't understand what I wanted, so everything they
> > suggested threw away resolution which we don't want to do, and gave
> > the same field of view in the output, but converted to useless fuzz.
> > Hence I come here in search of help.  Effectively, we need extreme
> > telephoto at maybe 57,600 pixels (240x240 for this example)
> > resolution, captured pixel by pixel from the central 57,600 pixels of
> > the input for this application.
> > 
> > Thank you for any usable suggestions.
> > 
> > Cheers, Gene
> 
> This is a job for a camera with a telephoto lens.
> You're using a camera with the wrong lens, that's all.

Several reasons not to.
1. Such lenses may be available where you are but just asking about them at 
stores where I can buy the camera gets me a "telewhat?"

2. Mechanical alignment, quite important, depends on the precision they use 
to mold the plastics.  And is generally non-repeatable.

3. Costs several times what the camera costs.

4, throwing away the unwanted pixels should be a few lines of code, 
essentially free, and by reducing the amount of data to be processed from 
5,000,000 pixels to 57,600 pixels gains me only .0868% times the data to 
process, which gains me many frames a second in processing speed, something 
your lens cannot not do.
> 
> For many cameras there are telephoto adaptors.

And how many can focus to 20mm from their front element?
 
> Capturing 2540x19xx pixels to get a 100x100 picture is
> an extremely unreasonable approach.
> 
Considering the cost of your alternative, its a most reasonable approach.
A run what ya brung approach to be sure, but if it gets the job done, who 
cares if I throw away roughly 95% of the picture?

Any body else?

Cheers, Gene
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)

"It's in process":
	So wrapped up in red tape that the situation is almost hopeless.
A pen in the hand of this president is far more
dangerous than 200 million guns in the hands of
         law-abiding citizens.


More information about the xorg mailing list