Operator-friendly machine vision tool for scrap identification
Title: |
Operator-friendly machine vision tool for scrap identification |
DNr: |
Berzelius-2023-303 |
Project Type: |
LiU Berzelius |
Principal Investigator: |
Sudhanshu Kuthe <kuthe@kth.se> |
Affiliation: |
Kungliga Tekniska högskolan |
Duration: |
2023-11-04 – 2024-06-01 |
Classification: |
20506 |
Homepage: |
https://www.kth.se/profile/kuthe |
Keywords: |
|
Abstract
With the need for energy-efficient steelmaking, the optimization of the EAF bucket
charge emerges as a crucial aspect. Effective scrap management and strategic
storage are pivotal for achieving sustainable steel production. In contemporary steel
plants, scrap is manually retrieved using mechanisms like the octopus claw and
magnets. For this system to function seamlessly, an advanced scrap identification
system is imperative. Such a system supports operators in the control room,
ensuring accurate scrap retrieval at designated intervals. This dual-objective system
aids in process control, ensuring precise scrap retrieval by the crane and
subsequently verifying the bucket layering needed for the EAF steelmaking.
An image database sourced from actual industrial enviornment is avaibale for data
processing. Leveraging image segmentation and object detection techniques, it
becomes feasible to classify specific types of steel scrap. The project proposes the
utilization of deep learning algorithms for these image processing tasks. The main
goal is to develop an operator-friendly machine vision tool for scrap identification.