2

I'm creating a set up that allows me to take photos of any object to capture its outlines for computer controlled cutouts. I'm no photography expert :D.

I use a 60 x 60cm LED Light panel to illuminate the surface to prevent shadows to get a clean picture, and a Canon 750D Kit (18-55mm Kit)

Adjusting some f-stop settings and shutter speed to get such an image. But i have a problem of distortion.

Using Adobe Illustrator, I will use Image Trace to get the outlines, to get an vector outline for a CNC router, and export as a Vector file. Whatever that is white, will be ignored. So technically speaking, i only need a silhouette shot :)

But i think I have a problem of Barrel Distortion, I'm not sure if this is a right type of distortion. Here is an image of a 'Samurai' Keychain from someone.

enter image description here

Notice the image near the sword handle (marked in a red square), It seems to look like a "diagonal/slanted" shot, which will cause a problem for me, because the outline will be thicker than the actual thickness of the guard, after image trace.

enter image description here

My question is: IS there any way or type of lens, at least not too expensive, to be able to take pictures of objects with the minimal or no distortion?

  • 1
    Using a zoom lense and shooting from far away will minimize the distortion ... if you really need to get rid of it, you could take multiple pictures while moving the camera from one side to the other, cut them in vertical slices and stitch them together – MoritzLost Mar 9 '17 at 19:49
  • Seems real tedious. Hmm, i'm just checking my options here. Yes it does feel less at more zoom. Thank you for your advice. Maybe we will just have to tolerate that – Paolo Mar 9 '17 at 21:40
  • The long focal length lens will probably be the easier way. Since you are only tracing the image, you don't need a good lens, just get a cheap 300 mm or something ... But if your only problem is the extended hilt and similar things, why don't you just manually fix the resulting vector graphic? Put the image below it on a preview layer and it should be easy – MoritzLost Mar 9 '17 at 21:43
  • 4
    You need to use a scanner, not a camera. – Michael C Mar 10 '17 at 3:46
6

In order to get an orthographic view of your subject using a camera, you would need a telecentric lens. One requirement for a telecentric lens is that the width of the lens be at least as large as the widest object you wish to photograph using it. If your Sword is three feet long, then you would need a telecentric lens at least three feet in diameter...

One solution many people desiring an orthographic view of an object adopt is to use a scanner, rather than a camera. The optical collectors in a scanner move linearly as they collect each line of information. This insures that a constant angle is maintained from one end of the scanned object to the other. One can also use a camera to take multiple images of an object while moving the camera linearly along the length of the object being imaged, then combining them in a linear or parallel motion panorama.

  • 2
    The scanner is the right solution. +1. Wanting to use the photo camera to scan an object is akin to the "when you only have a hammer, everything looks like a nail" common problem. – Olivier Dulac Mar 11 '17 at 10:07
5

That's not distortion. Your lens simply isn't directly over the handle. The camera can't be directly over every point on the sword, so you will always have this issue to a degree.

The best thing I can suggest is have as much working distance between the camera and subject: use a longer focal length to flatten/minimize the angles.

  • 2
    And in addition, if there is one part of the object that is more sensitive to this (such as the handle in this image), put that part in the middle of the image. – Roel Schroeven Mar 10 '17 at 8:26
2

Michael Clark's answer and MikeW's answer are both good ones and probably all you need, but I'll add another to the mix:

Instead of trying to capture the object itself, take a photograph of the object's shadow.

You say you only need the object's outline, and a shadow gives you exactly that. Shadows falling on a flat surface are inherently two-dimensional -- they have no thickness -- so photographing a shadow avoids the perspective problem that you're trying to avoid.

The main challenge in using a shadow is avoiding stretching and shearing distortions. To get a shadow that's the same size as the object, you'll want a light source with parallel rays, like the sun. If the surface on which the shadow falls is orthogonal (at right angles) to the light rays, and if the rays really are parallel, then the shadow will be the same size and shape as the object. However, in order to take an accurate photo, you'll need to put the camera in line with the light source and the object, so the camera's shadow will overlap that of the object. A solution is to have the light fall on the surface at a bit of an angle, which will elongate the shadow a bit, but to place the camera at the same angle in the opposite direction, which will cause an equal amount of foreshortening.

diagram

  • Very interesting indeed! Why haven't i thought of that. But I agree with you, the challenge of getting it all parallel is definitely a tough one. Nevertheless, I'll try when I can! – Paolo Mar 10 '17 at 10:01
  • Could also do it in transmission, with a partially transparent screen such as a sheet of paper. That way everything can be placed along the normal – baptiste Mar 12 '17 at 18:38
1

The other answers talk about some physical aspects of the setup, however, you may still encounter lens distortions bad enough to be a problem when cutting out your shapes. If there is a problem, there may be a solution.

No real-world lens will be perfectly free of any distortion effects. So the solution won't be to get a particular lens. Fortunately the solution is inexpensive, although the math can be a bit intimidating.

The answer will be to quantify the distortion of your particular lens and setup, and use that information to perform a mathematical transform of your images to square them out. The way you'd perform this function is to take a picture of a grid of known size and spacing. You then use some image processing algorithms such as those provided by OpenCV that can detect the presence of the grid, and measure the distortion in the image. You'd then use that information in the image post processing chain in preparation for your vector outline file.

This page http://docs.opencv.org/2.4/doc/tutorials/calib3d/camera_calibration/camera_calibration.html# has a broad view of the mathematical transforms and how they're used.

And detailed function parameter information is here http://docs.opencv.org/2.4/modules/calib3d/doc/camera_calibration_and_3d_reconstruction.html?highlight=fundamental that you would use to implement the image transforms in your application.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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