Precision Agriculture, 3, 63–80, 2002
© 2002 Kluwer Academic Publishers. Manufactured in The Netherlands.
Weed Detection Using Canopy Reﬂection
E. VRINDTS, J. DE BAERDEMAEKER, AND H. RAMON firstname.lastname@example.org
Katholieke Universiteit Leuven, Dept. AgroEngineering and Economics, Lab. AgroMachinery
and Processing, Kardinaal Mercierlaan 92, BLOK E, B-3001 Heverlee, Belgium
Abstract. For site-speciﬁc application of herbicides, automatic detection and evaluation of weeds is desirable.
Since reﬂectance of crop, weeds and soil differs in the visual and near infrared wavelengths, there is potential
for using reﬂection measurements at different wavelengths to distinguish between them. Reﬂectance spectra of
crop and weed canopies were used to evaluate the possibilities of weed detection with reﬂection measurements
in laboratory circumstances. Sugarbeet and maize and 7 weed species were included in the measurements.
Classiﬁcation into crop and weeds was possible in laboratory tests, using a limited number of wavelength band
ratios. Crop and weed spectra could be separated with more than 97% correct classiﬁcation. Field measurements
of crop and weed reﬂection were conducted for testing spectral weed detection. Canopy reﬂection was measured
with a line spectrograph in the wavelength range from 480 to 820 nm (visual to near infrared) with ambient
light. The discriminant model uses a limited number of narrow wavelength bands. Over 90% of crop and weed
spectra can be identiﬁed correctly, when the discriminant model is speciﬁc to the prevailing light conditions.
Keywords: weed detection, canopy reﬂectance, precision crop protection
For the optimal, site-speciﬁc application of herbicides, the weed population has to be
determined over the ﬁeld. This can only be done in an economic way by automating
the weed detection and evaluation (Wartenberg, 1996). The result of weed detection is a
weed map that can be used for determining the site-speciﬁc treatment of weeds. When
a weed map is recorded, weed detection and treatment can be performed at different
times and from different platforms. Another approach is on-line weed detection, where
detection and treatment are performed at the same time. An advantage of on-line weed
detection is the timeliness of the weed evaluation; the treatment can be adjusted to the
local weed pressure at a critical time for crop growth.
For site-speciﬁc crop protection, weed treatment is based on local weed populations.
The criterion for spraying could be a simple on/off control: spray when there is a weed,
do not spray when there is no weed. This strategy can yield a considerable reduction in
herbicide use for weeds that grow in patches and that are treated with speciﬁc products,
like grasses and thistle (Donald, 1994; Rew et al., 1997). Weed thresholds can also be
used in site-speciﬁc crop protection. Herbicide use will be determined by the possible
yield loss that the local weed population can cause. The possible yield loss can be
estimated when the type of weed and the weed density is known. A very important
factor in the yield loss caused by weeds is the relative age or growth of crop and weeds
(Cousens et al., 1987; Kropff, 1988). A high density of small weeds in a relatively older
and bigger crop will not cause a big reduction in yield, while a few relatively bigger
weeds can cause a substantial yield loss. The relative age of the crop and weed can be