Advice needed detecting counterfeit coins

[bmihura]

I’d like to use image processing to determine counterfeit coins by measuring their diameter and how circular they are.

My first attempt was to place the coins on graph paper with 1/4" squares, but ambient lighting often cast shadows. I found that displaying an image of those squares on my smartphone in a dimly lit room eliminates these shadows.

That being said, what image processing functions should I use, given the following:

1. For each picture I take, the camera will be held by hand so it may be a few degrees off and not directly perpendicular to the plane of the phone.

2. On my smartphone, the pixels are perfect squares, not rectangular. Their size and shape is the one constant I have to work with.

[muks]

---

@bmihura wrote:

I’d like to use image processing to determine counterfeit coins by measuring their diameter and how circular they are.

My first attempt was to place the coins on graph paper with 1/4" squares, but ambient lighting often cast shadows. I found that displaying an image of those squares on my smartphone in a dimly lit room eliminates these shadows.

That being said, what image processing functions should I use, given the following:

1. For each picture I take, the camera will be held by hand so it may be a few degrees off and not directly perpendicular to the plane of the phone.

2. On my smartphone, the pixels are perfect squares, not rectangular. Their size and shape is the one constant I have to work with.

---

What is the camera that you are using? Try using a high intesity light source and operate in mimum lens aperture.  Try thresholding and circle detection. Post a image is possible

[bmihura]

I'll be using a variety of cameras to develop this, but ultimately I'll write this as an Android app. To get the math right, I'll start by using NI's image processing software using your "thresholding" and "circle detection" suggestions. I've attached sample pictures where I use a common quarter.

[muks · Accepted solution]

Try this for the images you attached

[bruce008]

Nice! That worked perfectly, thanks.

[muks]

Dont forget to mark the post as solution!!!

[bmihura]

That worked! Still not sure how to accept your solution though, will hunt around until I get it....