Machine vision in transit quality control: Automated inspection of pulp bale loading

Thumbnail Image

Files

master_Chaturvedy_Anushka_2025.pdf (21.76 MB)

URL

Journal Title

Journal ISSN

Volume Title

School of Science | Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.

Authors

Date

2025-07-31

Department

Major/Subject

Machine Learning, Data Science and Artificial Intelligence

Mcode

Degree programme

Master's Programme in Computer, Communication and Information Sciences

Language

en

Pages

77

Series

Abstract

As global trade expands at a rapid pace, ensuring product quality is not compromised during transportation is vital for businesses. In the pulp and paper sector, transit quality control (TQC) implementation is done manually in most cases, and the possibility for unsafe conditions and shipment damage is significant. The same applies to UPM Pulp, the case company. This Thesis presents a lightweight machine vision pipeline that aims to automate the measurement of an important quality control parameter for TQC, where errors have been identified to occur most frequently during field audits, and can lead to significant damage during shipment. The Thesis focuses on UPM Pulp’s North American logistics operations. The study aims to evaluate the proposed machine vision pipeline combining object detection with either of the two detection heads, YOLOv11n (You Only Look Once version 11 nano) or Azure Custom Vision (ACV), with a pretrained segmentation model, called Segment Anything Model (SAM). Results are compared against ground-truth measurements. The study compares two annotation methods (bale-by-bale vs stack-level) and two measurement approaches (Mean-of-two vs Airbag width), by deploying an OLS regression model along with error metrics. The best performing scenario achieved a mean absolute error (MAE) of 3.46 cm. This study presents a proof-of-concept solution to automate transit quality checks, using images captured on-site in a non-standardised setting, and a lightweight machine vision model, which could enable adherence towards safety and compliance standards, as well as reduced damage to the shipment, hence positively impacting the pulp business and supply chains.

Description

Supervisor

Kannala, Juho

Thesis advisor

Kumorovitzova, Andrea
Kaki, Anssi

Keywords

machine vision, computer vision, segment anything model, transit quality control, automation, you only look once

Other note

Citation

Permanent link to this item

https://urn.fi/URN:NBN:fi:aalto-202508196338

Collections

[dipl] Perustieteiden korkeakoulu / SCI