hh.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • 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
Crop and weed discrimination using computer vision
Chalmers University of Technology.
2006 (English)Licentiate thesis, comprehensive summary (Other academic)
Abstract [en]

The thesis is concerned with computer vision in ecological and precision agriculture aiming to identify crops and weeds in images of crop rows. The identification of plants can be done by extracting the plants in the image and classifying them as a crop or weed depending on the shape and color. The extraction of plants can be accomplished in two ways, either by extracting the separate leaves and combining them into plants or by extracting the plants directly. Plant overlapping each other, plants missing parts of leaves or whole leaves, variations in plant appearance, and missing plants in crop rows are the main problems the computer vision-based plant identification needs to tackle.

The main objective in this thesis is to develop and investigate methods for both ways of extracting the plants and classification of extracted plans into crop or weed classes. A new method for extraction of separate leaves, called cutting, is presented and compared to the watershed and erosion followed by dilation methods, which have been used in other applications of leaf extraction. For the direct extraction of plants, the active shape models (ASM) method is adapted and evaluated. The initial position is one of the main prameters affecting the robustness of the result obtained from the ASM. Therefore the robustness regarding different initial positions is studied experimentally.

The experimental investigations are performed on images of crop rows taken from different ecologically grown sugar beet fields, including images of both overlapping plants and plants missing part of leaves or whole leaves. The investigation results show that the cutting method produces crop leaf segments that resemble the leaves with about 80% average accuracy and removes 61% of the occluding weed pixels. The active shape model method removes up to 83% of the occluding weed pixels and categorizes up to 83% of the crop pixels correctly. Arround 84% of the plants extracted using the active shape model were correctly classified. This is fairly close to what can be achieved by manual extraction of the plants. The active shape model robustness test regarding the initial position shows that the plant is extracted properly if the model is placed within half a leaf distance from center, mosrotated by at most +/- 18.5 degrees and the scale parameter of the model does not exceed twice the plant size.

Place, publisher, year, edition, pages
Göteborg: Chalmers university of technology , 2006. , p. 44
Series
Technical report R, ISSN 1403-266X ; R014/2006
Keywords [en]
Image analysis, Crops, Weeds
National Category
Other Computer and Information Science
Identifiers
URN: urn:nbn:se:hh:diva-1979Local ID: 2082/2374OAI: oai:DiVA.org:hh-1979DiVA, id: diva2:239197
Presentation
(English)
Available from: 2008-09-29 Created: 2008-09-29 Last updated: 2018-03-23Bibliographically approved

Open Access in DiVA

No full text in DiVA

Authority records BETA

Persson, Maria

Search in DiVA

By author/editor
Persson, Maria
Other Computer and Information Science

Search outside of DiVA

GoogleGoogle Scholar

urn-nbn

Altmetric score

urn-nbn
Total: 136 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • 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