Abo exklusiv

Wissenschaftliche Zeitschriften

Mehr über das wissenschaftliche Verlagsprogramm erfahren Sie im Zeitschriften-Überblick im » Ulmer-Zeitschriftenkiosk

Image analysis for automatic classification of Rumex obtusifolius in mixed grassland swards

Summary
The complete article is written in German
Recent developments in site-specific weed control in arable crops are using digital image analysis for plant species identification. In grassland, these techniques have rarely been applied so far. The presented investigation deals with the development of methods for automatic weed detection in grassland based on digital RGB images. The focus was to detect Rumex obtusifolius (RUMOB) and other accompanying herbs, like Plantago major (PLAMA) and Taraxacum officinale (TAROF). 24 bit RGB images were transformed to 8 bit intensities. Based on that the local homogeneity was calculated from which a binary image was derived. Morphological opening was applied finally. The result was, that leaves of the weeds could be segmented from the image background. For each of object colour, texture and geometry features were calculated. Based on these features a Maximum-likelihood Estimation (MLE) was calculated in order to classify the objects into the classes (i) RUMOB, PLAMA, TAROF, soil, residue und (ii) RUMOB und residue. The average RUMOB detection rates ranged from 70,9 % to 95,3 %.

german version

Keywords/Stichworte:Weed mapping; digital image processing; pattern recognition; Rumex obtusifolius

Downloads: (Hinweis zu PDF-Dokumenten)

Download (Full Article in german) (PDF)

Autor:S. GEBHARDT, W. KÜHBAUCH, Institut für Pflanzenbau, Universität Bonn, Katzenburgweg 5, D-53115 Bonn, E-Mail: s.gebhardt@uni-bonn.de, lap@uni-bonn.de

Drucken Weiterleiten