0
\$\begingroup\$

We're trying to figure out how to capture a device that has a rectangular shape of 1.4m x 0.4m.

It is made of 4 lanes of 1.4m x 0.1m with walls which are 0.07 tall.

We'll need to build a video stream of it, so only real-time solutions are acceptable, and need to produce a digital output that is as distortion free as possible (which is a problem since it's 'full of long lines' and distortion is readily visible).

We have been discussing the following:

  • get a lens that has a narrow FoV and put the camera far.
  • get a lens with a wide FoV and correct the image through real-time processing.
  • put several cameras side by side and do real-time stitching, although this will probably also require correction for the distortion

Since none of us really know that field, I'm wondering if anyone here would have some solution to recommend.

Improve this question
\$\endgroup\$
14
  • 5
    \$\begingroup\$ While I also do prefer that someone offers feedback on how to improve when downvoting, it is not, in fact, never cool to down vote without explaining. This has been discussed extensively on meta.SE and there is a reason why the system allows downvoting without requiring a comment first. Like it or hate it, downvoting without explanation is accepted behavior on SE. \$\endgroup\$
    – AJ Henderson ♦
    Commented Sep 19, 2017 at 14:11
  • 3
    \$\begingroup\$ That said, my best guess about the reason for the down vote (I'm not the DVer) is that the question doesn't seem particularly photography oriented. It sounds more like a scientific instrumentation question and particularly seems like it is more video oriented than photo. It's still arguably on the edge of on topic since the optics questions are still possibly slightly relevant to still photography as well, but that may be the reason. The question itself is also a bit hard to parse what you are trying to accomplish. \$\endgroup\$
    – AJ Henderson ♦
    Commented Sep 19, 2017 at 14:13
  • 2
    \$\begingroup\$ The thing that makes it borderline is that while lenses and optics is on topic as relevant to photography. Having industrial levels of lack of distortion isn't really typically a photography concern. Sure less distortion is usually better than more, but making determinations about lens choice relative to a particular level of needed precision is not. You are looking for stuff that is more typically a measure in scientific instrumentation rather than photography. It might still be able to get an answer here, but it feels very borderline to the topic of the site. \$\endgroup\$
    – AJ Henderson ♦
    Commented Sep 19, 2017 at 14:46
  • 6
    \$\begingroup\$ I'm voting to close this question as off-topic because while the question may be borderline overall, the constraint "We'll need to build a video stream of it, so only real-time solutions are acceptable" takes it squarely out of scope. There is no intent to produce a photograph here, and you've ruled out answers that would be relevant in a photographic context. \$\endgroup\$
    – mattdm
    Commented Sep 19, 2017 at 17:09
  • 1
    \$\begingroup\$ @matt: it is a lens / processing question; whether I take one photo or 30 photos per second is not really relevant but for the processing part \$\endgroup\$
    – Thomas
    Commented Sep 19, 2017 at 17:30

3 Answers 3

Reset to default
1
\$\begingroup\$

You should use a telecentric lens. For your application, given the size of your target object, this isn't a cheap option but it's the most appropriate given the specs you've provided.

Improve this answer
\$\endgroup\$
2
  • 1
    \$\begingroup\$ Note that telecentric lenses are fixed-focus. They are also referred to as f-theta lenses. \$\endgroup\$
    – Stan
    Commented Oct 27, 2017 at 4:41
  • 1
    \$\begingroup\$ From that Wikipedia page: "Telecentric lenses tend to be larger, heavier, and more expensive than normal lenses of similar focal length and f-number. This is partly due to the extra components needed to achieve telecentricity, and partly because the object or image lens elements of an object or image-space telecentric lens must be at least as large as the largest object to be photographed or image to be formed." If I understand that correctly, the lens should 1.4 m in diameter; doesn't really sound practical. \$\endgroup\$
    – Roel Schroeven
    Commented Oct 27, 2017 at 8:17
1
\$\begingroup\$

Geometric distortion is an easily measurable characteristic of objective and is available for almost every objective ever made. There is a huge collection of optical measurements for modern objectives: DXOMark.com, OpticalLimits.com, Imaging Resource and others. Here is what they say about distortion:

Zooms usually have negative distortion (barrel) for short focal lengths, and positive (pincushion) distortion for longer focal lengths. The metric penalizes both types of distortion. Distortion is expressed as a percentage: the value 0 is the perfect case; 1% is high, but there is no upper limit. A value of 0.2% corresponds to a noticeable distortion. Wide-angle lenses have more distortion.

https://www.dxomark.com/About/Lens-scores/Metric-Scores

The positional error measurement related to this series of points is the ratio 1/L, expressed as a percentage (since it is a distance ratio)

https://www.dxomark.com/About/In-depth-measurements/Measurements/Distortion

There is always some extent of distortion you can tolerate - it may be one pixel, for example. One pixel tolerance for 1280 pixel wide frame is 0,07% in DXO terms. And there is a plenty of objectives to choose from:

https://www.dxomark.com/best-lenses-under-13000-dollars (sort them using distortion column)

DXO does not measure distortion smaller than 0,1% or so (but OpticalLimits has more fine measurements). Additionally you can improve distortion characteristic by using only part of image circle (i.e. using Micro43 camera for a 135 frame objective will improve said percentage roughly two-fold).

Improve this answer
\$\endgroup\$
0
0
\$\begingroup\$

In my experience with gaming, it is best to avoid any image processing or change to original data that can give rise to suspicions of veracity. There should be some means of ensuring impartiality among the discrete events.

I have the same opinion about multiple cameras so long as any data is not redundant for the event.

My experience is with horse racing. My education is in imaging science. My interest is psychology.

I would choose a point of view directly above the "finish line" to capture the result of the competition.

I would choose a point of view directly above the "fifty-yard line" to capture the relative progress of the competitors.

I would choose a point of view directly above the "start line" to capture the beginning of the event.

In addition to sensors, a calibrated track should also used to aid the interpretation of the competitors' relative progress during the event. The concept is similar to the mile markers used on some US highways to aid aerial spotters to determine illegal car speeds. Lines, dots, squares, anything to use as a reference mark that is the same among the competitors will be suitable.

The kind of lens should be considered in addition to focal length/field of view. The accurate linear rate-of-change of the image across the image plane might call for a linear "f-theta" lens. (Pictorial lenses are "f-sine-theta" lenses.)

EDIT: For human consumption, I would use a split screen showing multiple camera angles. I would use an overhead view, an eye-level oncoming view, and an oblique "over the shoulder view" of the finish line as seen from the start.

Second EDIT: Take some time to verify that your device can be calibrated with a "standard" competitor to give consistent, unbiased, and consistent results.

Place your bets!

Improve this answer
\$\endgroup\$
3
  • \$\begingroup\$ multiple split screen seems like a great idea! but the original problem stands: the track is quite long and distortion is quite visible \$\endgroup\$
    – Thomas
    Commented Sep 21, 2017 at 13:37
  • \$\begingroup\$ Can you make the distortion work for you? So long as the image highlights the "money shot" in a manner that minimizes opportunity for any dispute, it should be good-to-go. A photo finish is distorted to enlarge and freeze an image of the noses of the competitors relative to the finish line. The question to ask of any solution is if it is acceptable. \$\endgroup\$
    – Stan
    Commented Sep 22, 2017 at 1:26
  • \$\begingroup\$ The result comes from sensors anyways, not from the video, so the results are safe; I don't know if a more 'artistic' view from the cam would work, I'll ask! \$\endgroup\$
    – Thomas
    Commented Sep 24, 2017 at 10:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.