フィルターのクリア

Detection, labeling and area calculating of glass plate fragments

2 ビュー (過去 30 日間)
Ezzat
Ezzat 2022 年 1 月 30 日
コメント済み: Image Analyst 2022 年 2 月 5 日
Hi All,
I do a fragmentation test for a laminated glass plate.
The shown image illustrates the glass plate after impacting it such that there are no separation between fragments after impact.
I'm aiming at detection, labeling and then area calculating of the fragments in the whole image with help of the shown caliper value.
Thank you

回答 (1 件)

Image Analyst
Image Analyst 2022 年 1 月 30 日
First, I'd try using a telecentric lens. You are not using one in this image and that is why the edges are more contrasty near the optic axis and blurrier away from it. A telecentric lens will eliminate that problem.
Next, try different backgrounds - both black and white and see which gives you greater contrast.
In addition to that, try different lighting geometries. What do the cracks look like with broad overheard lighting? How about line sources of light (LED light strips/bars) down inthe plane of the glass pointing inward? How about two point sources of light at 45 degree angles shining down at a 45 degree angle?
If all goes well you'll have either dark or bright cracks against a uniform background and you can simply use thresholding. But don't forget to take a uniform image of the background so that you can divide out the background to flatten the image and get rid of lens shading and light non-uniformities.
As you have it now, I think no matter what you do you won't have satisfactory results. There's just no way some/most of the cracks or solid domains between them can be resolved. Of course some of the image you can get them, but not over the whole image.
So fix your image capture conditions and then come back with your improved image. If you don't know how, then contact a reputable machine vision vendor (not just someone who sells parts but someone who actually builds turnkey systems).
  3 件のコメント
Ezzat
Ezzat 2022 年 2 月 5 日
I put the scanned pictures but I didn't have any feedback ??
Image Analyst
Image Analyst 2022 年 2 月 5 日
Try using watershed.

サインインしてコメントする。

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by