1

I'm working on various methods to count objects on conveyor belt using camera feed. Methods used:

  1. OpenCV Background subtraction
  2. OpenCV Color Tracking of objects in the video

The main issue faced is that even with a small change in the environmental conditions affects the detection of objects and sometimes misses the object completely. Envronmental conditions include light conditions when there is too much brightness or too much dark.

Please suggest any improvement or any other methods to solve this issue Also trying the tensorflow object detection model to detect the object. I want to have 100% accuracy in object counting in any condition

Gaurav Sisodia
  • 31
  • 5
  • 3
    impossible to discuss without source data and your current code and results – Christoph Rackwitz Apr 20 '22 at 13:01
  • Since the objects are moving over a conveyor belt in one direction, assume a line which the contour of the object passes. Once it crosses the line you can increment the counter. This is just a simple approach but you can try – Jeru Luke Apr 20 '22 at 20:08
  • Thats the core logic for counting the objects, when using OpenCV background subtraction, but there is the issue that it detects other objects like shadow, reflection, unwanted material on conveyor belt, hence accuracy decreases. That's why is used OpenCV Color Tracking to track color of objects for counting, but due to variable changes in light conditions , the color changes hence missing the object completely. @Jeru Luke – Gaurav Sisodia Apr 21 '22 at 05:17
  • @GauravSisodia For tracking colored objects you can try converting the frame to a different color space (LAB, HSV, etc.) and analyze the individual channels – Jeru Luke Apr 21 '22 at 06:00
  • Can you provide me some link or some articles to know more about these different color space, btw i was using RGB, but if LAB or HSV provides different results , i'm willing to use them. Thank you @JeruLuke – Gaurav Sisodia Apr 21 '22 at 08:26
  • @GauravSisodia Pls check the following answers https://stackoverflow.com/questions/71944027/unable-to-read-captcha-with-python-tesseract-and-opencv/71944923#71944923 and https://stackoverflow.com/questions/71818490/how-to-isolate-yellow-part-of-the-picture-and-make-black-other-parts/71824363#71824363. In the first case I have used HSV color space while in the second case I have used LAB color space – Jeru Luke Apr 21 '22 at 08:35
  • Also check out https://stackoverflow.com/questions/71654740/how-to-remove-a-watermark-from-a-document-image/71656217#71656217 to select a range for your color of choice – Jeru Luke Apr 21 '22 at 08:38

0 Answers0