liu.seSearch for publications in DiVA
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • oxford
  • 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
Cross-Modal Detection and Matching of Wood Logs: Object Linking Between RGB and Laser Range Data
Linköping University, Department of Electrical Engineering, Computer Vision.
Linköping University, Department of Electrical Engineering, Computer Vision.
2025 (English)Independent thesis Advanced level (degree of Master (Two Years)), 28 HE creditsStudent thesis
Abstract [en]

The adoption of computer vision and machine learning in industrial settings offers significant potential for improving productivity and efficiency. One way to realize these benefits is through methods for connecting data across different sensing modalities. This thesis investigates the detection and cross-modal linking of individual wood log ends across RGB images and laser data. RGB images capture visual characteristics such as wood type and quality, while laser data provides volumetric measurements. By linking logs between these modalities, it becomes possible to extract a more complete and detailed description of each log, thereby supporting automated inventory and quality assessment.

To achieve this, two methods were proposed. The first involved using a pre-trained feature matching model to establish correspondences between the modalities, leveraging patterns and features common to both. The second method focused on calibrating the RGB cameras to align and map the RGB images onto the laser data. For this approach, a segmentation model was trained on various types of laser data to detect individual logs.

The results revealed that the pre-trained feature matching model lacked sufficient generalization to handle the cross-modal data used in this thesis. However, training the model on the cross-modality data could potentially enhance its performance. In contrast, the second method demonstrated strong potential, achieving an 86% success rate in linking individual logs, as evaluated on a test dataset with manually annotated ground truth.

Place, publisher, year, edition, pages
2025. , p. 70
Keywords [en]
Object Linking, Feature Matching, Cross-Modal Matching, Instance Projection Log Matching, RGB and Laser Data Linking
National Category
Computer Vision and Learning Systems
Identifiers
URN: urn:nbn:se:liu:diva-214522ISRN: LiTH-ISY-EX--25/5736--SEOAI: oai:DiVA.org:liu-214522DiVA, id: diva2:1966314
External cooperation
Mabema AB
Subject / course
Computer Vision Laboratory
Presentation
2025-05-28, Visionen, Stora konferensrummet, 15:15 (English)
Supervisors
Examiners
Available from: 2025-06-16 Created: 2025-06-10 Last updated: 2025-06-16Bibliographically approved

Open Access in DiVA

fulltext(170645 kB)202 downloads
File information
File name FULLTEXT01.pdfFile size 170645 kBChecksum SHA-512
32ba507507ec95a98ac932be4a0855835b0185c080b123ab7eeb5e05c8aff44e2043fe493283603567414cd423fbd0562c21500249c440292ef8250722281a8d
Type fulltextMimetype application/pdf

By organisation
Computer Vision
Computer Vision and Learning Systems

Search outside of DiVA

GoogleGoogle Scholar
Total: 202 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 398 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • oxford
  • 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