Weed identification using digital image analysis

H. Oebel, R. Gerhards, M. Sökefeld, P. Risser, M. Pflugfelder, A. Nabout, W. Kühbauch

Introduction

Image analysis and classifikation

Image analysis
Process of image analysis
Comparison to database
Comparison with database depending on decision criteria

Creation of application maps

Documentation of weed distribution for three weed classes
Distribution of weeds
Distribution of broad-leaved weeds and grasses in sugar beets
Distribution of weeds
Distribution of broad-leaved weeds and grasses in malt barley

Results of the automatic identification process

% Identification Maize CHEAL ECHCG SOLNITotal
Maize 90 0 2 0
CHEAL 0 94 19 5
ECHCG 4 5 80 17
SOLNI 6 1 1 78
Total86

Recognition rate of the digital image analysis in Maize
% IdentificationSugar beetDicotsGrassesGALAPTotal
Sugar beet80785
Dicots97984
Grasses886914
GALAP351477
Total77

Recognition rate of the digital image analysis in sugar beet
% IdentificationWinter wheatGrassesDicotsGALAPTotal
Winter wheat80000
Grasses1310000
Dicots008111
GALAP701989
Total88

Recognition rate of the digital image analysis in winter wheat
% IdentificationMalt barleyDicotsGrassesGALAPTotal
Malt barley699913
Dicots97885
Grasses1067410
GALAP127972
Total 73

Recognition rate of the digital image analysis in malt barley

Conclusion

The classification algorithm is suitable for the identification of weed species.