Classification of Meat-Grinder Chucks Using Image Processing

EasyChair Preprint 6638

Abstract

The work performed includes an end-to-end solution to a product classification problem with machine vision. Meat-grinder chucks separated by manpower within the company cause loss of time and errors. To solve the problem, chucks are detected with an infrared sensor, and the photos are taken. First the outer diameter of the chuck is calculated by Hough transform, after that, focusing inside the chuck, the radii of the inner circles are found. The circles found (in order to evaluate more accurately) are trained via a machine learning algorithm Naive Bayes method, the model is obtained. It has been determined in the study that this problem can be solved by a simpler method, mode operation. The output of the trained model indicates which class the product belongs to. The pneumatic pistons in the system are triggered by the time cutter software running on it and the products are separated to determined box. As a result of the tests performed on the system with the method of taking mode, it was determined that the model works with 99.2% accuracy. In the tests performed with the Naive Bayes method, an accuracy of 95.8% was achieved.

Keyphrases: Automation, Industry 4.0, image processing, machine vision

@booklet{EasyChair:6638,
  author    = {Muhammed Başarkan and Metin Turan},
  title     = {Classification of Meat-Grinder Chucks Using Image Processing},
  howpublished = {EasyChair Preprint 6638},
  year      = {EasyChair, 2021}}