Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Automation of quality inspections of recycled wood
University West, Department of Engineering Science.
2023 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

This research thesis explore the implementation of machine vision and artificial intelligence in automating the process of rating the quality of recycled wood. The goal is to contribute to the circular economy by promoting the reuse of wood in the construction industry. The thesis will focus on detecting visual defects such as nails, knots, and cracks on the surface of wood planks using a machine vision system integrated into a mobile station. It also includes a cost analysis of the required hardware and software. The research will utilize software tools such as Visual Studio code with Python 3.10.9 and Google Colab for training the AI model. Key hardware resources include a conveyor, wood samples with different qualities and defects, Raspberry Pi 4, Oak D lite camera, SD card, and a 5V power supply.The method described in this research study uses a machine vision system with a customized object identification neural network to find defects on the wood planks and measure them. Five distinct planks were used to test the object detection model, and the results were reported in a table. The model precision was discovered to be 72.6%. The neural network was shown to have a 67.0% accuracy rate. The model recall, which gauges how accurately it detected defects, was found to be 89.8%. The difficulty of identifying various defect types due to variances in appearance and imbalance in the training data is discussed in the study. The measurement of plank dimensions using machine vision methods is also evaluated in the paper. Five planks had their length, width, and depth measured, and the absolute and relative errors were computed. The width measurements had an acceptable average absolute error of 0.84 mm and a relative error of 0.92%. There was space for improvement in the depth measurements, which had an average absolute error of 1.8 mm and a relative error of 7.98%. The length measurements, on the other hand, revealed a higher absolute and relative inaccuracy, pointing to measurement algorithm limits. It is demonstrating the possibilities of employing AI for measuring and detecting defects in wood planks. It recognizes the length measurement as an area that needs further work and emphasizes the need for a more consistent dataset to increase the accuracy of the detectionproblem. The constraint of not taking into account the internal defects of the planks is also highlighted in the report, which also underlines the requirement for standards in recycled wood quality inspection for the advancement of circular economy activities.

Place, publisher, year, edition, pages
2023. , p. 35
Keywords [en]
machine vision, AI, wood
National Category
Robotics
Identifiers
URN: urn:nbn:se:hv:diva-20389Local ID: EXC915OAI: oai:DiVA.org:hv-20389DiVA, id: diva2:1780140
Subject / course
Robotics
Educational program
Master i robotik och automation
Supervisors
Examiners
Available from: 2023-07-17 Created: 2023-07-05 Last updated: 2023-07-17Bibliographically approved

Open Access in DiVA

No full text in DiVA

By organisation
Department of Engineering Science
Robotics

Search outside of DiVA

GoogleGoogle Scholar

urn-nbn

Altmetric score

urn-nbn
Total: 224 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf