0

I am new to OpenCV with linux. I am currently working on tube rack's bounding detection. But as you can see, one of rack's bounding is blurred by shadow.

enter image description here

Despite having used histogram equalization to enhance the contrast, there is still significant edge loss, which has greatly affected my edge detection.

enter image description here

I hope someone can give me some suggestions. Thank you very much.

I tried increasing the contrast and lowering the Canny detection thresholds, but I still couldn’t achieve good results.

Improve this question
5
  • 2
    adjust the lighting a bit so that the tubes are darker. Go with colour thresholding and contour detection. A bit of hardware adjustment will do wonders that software might not be able to do Commented Sep 8 at 7:39
  • 2
    Marking the corners of the tray in colour(s) might be an idea. Or using a more contrasting background on the workbench. And using a longer lens positioned further away would give less distortion and rounder, more overhead views into the contents of the individual "pots". Commented Sep 8 at 9:52
  • 1
    stop trying to make canny work for this. it won't work. discard that line of thought. -- in previous such questions, marking the corners and detecting them, then considering a homography for the tray's surface plane, also was the most promising approach. Commented Sep 8 at 10:24
  • 2
    there's no point in detecting individual circles. if you did, then you'd be left with a set of circles, and what good is that gonna do? drawing circles on the circles is obvious but useless. you probably need the lattice of them, and reconstructing a lattice from a set of points is more work than simply detecting the entire tray and then figuring from the tray to the individual holes. you have foreknowledge of where the holes are in relation to the tray. Commented Sep 8 at 10:29
  • 2
    go find retroreflective sticky film/foil that is already cut into round pieces (or get someone to cut them for you, or use a two-hole punch for office work). the retroreflectivity will make all the contrast/illumination problems go away. stick four them on the corners, precisely. stick a fifth close to whatever corner you want to designate for orientation. the rest is exercise. Commented Sep 8 at 10:31
Related questions
0
How can I detect all objects in this image with a white background without merging them with each other?
0
openCV Canny edge detection improvement
423
How to crop an image in OpenCV using Python
Load 2 more related questions Show fewer related questions

0

Reset to default

Your Answer

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

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.