Re: Measure distances w 1mm precision

Hi

As I know a sensor with 10000 x 10000 Pixel is not available. I know a sensor with 13 000 x 7 000 and this cost approx "only" 20 000 Euro, of course without housing and lenses :-).

As you can see big sensors are not really cheap. For a good camera system you need at least 3000-5000 Euro (camera and lenses). With a 20 mm lenses and a sensor with 1280 x 1024 pixel (6.7 um Pixel Size) you will have an field of view of 850 mm at a distance of 2 m. Lenses with a smaller focal length are also really expensive and not accurate enough for measurement systems.

Yes, you can measure object positions better than 1 pixel in a image. The easiest one is to have a round target. There exists a lot of algorithm to find them in a image and to measure it accurately. A other possibilities is to use least square matching (template matching).

If you want more details search for "close range photogrammetry". There are lot of people working on field.

Regards
Jürgen


googlegroups.9.multiversum4@xxxxxxxxxxxxxxx schrieb:
Thank you very much for your answer, Juergen!

Some things I don't quite understand:

What do you mean with 1/10 pixel measurement accuracy? Mustn't 1 pixel
per definition be the smallest unit of measurement accuracy? And why do
you use 0.1 mm in your answer when I think 1 mm would be good enaough?
Of course, if the object is white and the background black, then some
interpolation might be done from the grayscale of the pixels on the
edge of the object. Is that what you mean?

I've just been reading some introductory texts on camera calibration,
and it seems to be a subject rich with technics to make wonders.

My main concern is now which kind of camera I would need. What
parameters should I look for? I really don't know much about cameras.
Since I cannot choose the distance freely, I suppose that I need to
choose optics able to focus on the area of interest (i.e. about 1 meter
wide at 2 m distance) and with a sensor of enough precision and size.
Since 1 m is 10 000 times the 0.1 mm resolution you calculated with,
Juergen, would it be correct to infer that the sensor should be 10 000
* 6 um = 60 mm wide in order to capture a 1 m wide area at 2 m
distance?

What I want to do is make real time prognos of when the moving object
will touch the larger fixed edge (max 30 or so mm away). It probably
accelerates fairly regularly. Once its behavior has been studdied and a
model for its movement has been derived, repeated photos and
calculations on them could probably give good prognosis with lesser
resolution than 1 mm.

Best regards!


Juergen Hefele skrev:

Hi

Before you start you have to some estimation about the focal length of
your camera. Let's say you have a measurement accuracy of 1/10 Pixel in
your image (maybe a little bit optimistic) and your distance to the
camera is 2 m. The you get with intercept theorem:

f=focal length
p=distance on the sensor (e.g 1/10 Pixel of 6 um (typical pixel size on
the sensor)
d=distance camera object (here = 2 m)
x=distance on the surface (you want to measure at least = 0.1mm)

f/p =d/x

then you need at least a focal length of 12 mm. Be careful this
approximation is only valid in the mid of the image and it's really
optimistic! At the border the accuracy is not so high. With the given
sensor size you can calculate the field of view (Same formula as above
p=length of the sensor)

To get such a high accuracy you need a very good camera calibration.
Maybe you can use the approach from Tsai (or the one which is
implemented in OpenCV) But don't use Zoom lenses you don't get this high
accuracy with them.

It must be also clear that the distance camera and your objects must be
well known and stable or you must calibrate the system before each
measurement.

But with all of them you will get such a high accuracy.


Regards
Jürgen

.

Relevant Pages

  • Re: Megapixel War Over for Point and Shoots?
    ... small sensor and the contrast detect focusing. ... you can get a good D-SLR with two decent, image-stabilized, lenses. ... P&S cameras can have more seamless zoom range than any DSLR glass in ... slightly greater detail retention when upsampled directly in the camera for JPG ...
    (rec.photo.digital)
  • Re: The "Oh no not another Burpday Moot" Thread
    ... it's not a bad camera. ... and the bigger the physical size of the sensor relative to the ... Cheaper/smaller lenses usually let less light through (the maximum ... Bigger sensor elements means more sensitivity (but lower sensor ...
    (uk.rec.sheds)
  • Re: DPI
    ... The pixel originates in your camera. ... There are roughly eight million Pixels on the sensor in your ... resolution discernable by the human eye. ...
    (alt.photography)
  • Re: The "Oh no not another Burpday Moot" Thread
    ... I'm still looking for an easy to use fingery camera that will take ... and the bigger the physical size of the sensor relative to the ... Cheaper/smaller lenses usually let less light through (the maximum ... Bigger sensor elements means more sensitivity (but lower sensor ...
    (uk.rec.sheds)
  • Re: Any thoughts /news on Foveon sensors?
    ... >the design of the sensor and the fact that the pixel count hasn't changed. ... The X3 sensor doesn't suffer from colour moire on B&W images, ... >anti-alias filter of a Bayer sensor camera. ...
    (rec.photo.digital)