-1

I have image with wood trunks.

I have to detect each wooden trunk individually. It looks similar like following image:

wooden trunks

Do you have any ideas about approaches how to do that?
Should I use Al? Or just machine learning like SVM? Or some pattern recognition algorithm? Or I can train it.

training dataset

I tried to detect circles/ellipses, but it doesnt have good results. I also read that wood reflect red color.
But I dont have so much experience with OpenCV, so I dont know which approach is the best for this task.
Thank you for your help

user12651635
  • 29
  • 3
  • 1
    If you have enough data, YOLO object detector would be good for that. Besides with suitable thresholding and hough transform you can get a good results too (opencv). – MeiH Jan 04 '20 at 10:30
  • Thanks for your advice. I will look at it :-) – user12651635 Jan 04 '20 at 11:35

1 Answers1

0

I think retraining YOLO seems like a good option:

https://github.com/AlexeyAB/darknet

You'll need about 2,000 labeled images, plus image augmentation. I've used this library for image augmentation for YOLO:

https://github.com/aleju/imgaug/

Derek Pankaew
  • 64
  • 2