Vol. 5, No. 2, 173-179, 2006

Classification of weed patches in QuickBird images: Verification by ground truth data
Matthias Backes and Jan Jacobi

Current methods for mapping weeds in arable land include manual sampling approaches and online computer-based methods with special sensors. Both methods are expensive, time consuming and not suitable for constructing regional maps of weed status. This study investigated the use of a visual interpretation of high-resolution satellite images from the QuickBird satellite in order to detect weeds in a field of sugar beets (Beta vulgaris L.) near Bonn in Germany. The study compared this visual interpretation with the data acquired applying a WeedScanner survey of the same area. This method allows an exhaustive survey of weeds in the field. The analysis showed that dense clumps of Canada thistle (Cirsium arvense L.) were accurately detected in the satellite images, but that small and sparsely occurring weeds could not be reliably detected. The results prove the limitations of remote sensing in the context of weed control but they also show that there is a great potential for early decision making for particular weed species.

View Full Text (pdf file, 1.5 MB) previous page
Submitted: 25 May 2004
Revised: 28 June 2006
Accepted: 28 June 2006
Published: 03 July 2006

Backes M & J Jacobi, 2006. Classification of weed patches in QuickBird images: Verification by ground truth data. EARSeL eProceedings, 5(2): 173-179


EARSeL European Association of Remote Sensing Laboratories, Strasbourg, France


BIS Library and Information System, Carl von Ossietzky University of Oldenburg


ISSN 1729-3782