Agriculture Reference
In-Depth Information
of the field where crop parameters, estimated via local perception, are represented
according to the instantaneous position at which each measurement was carried out.
The assessment of crop status with local perception offers a practical and afford-
able alternative to remote sensing. Satellite imagery cannot compete in resolution
and updating rate with on-vehicle perception.
Thermographic maps
are generated
by special cameras that are sensitive to infrared, producing 2-D images that repre-
sent variations in temperatures. These variations can be spatially referenced to the
field in order to highlight points of extreme values in temperature, and correspond-
ingly in heat. Thermography has been used to estimate the
water content
in the
field, and as a means to quantify
water stress
—a factor indicating grape quality—in
wine grapes. Additionally, thermal cameras have been integrated in safeguarding
systems of autonomous vehicles to detect living beings interfering with the vehicle's
trajectory and unnoticed by sensors with problems to penetrate vegetation.
Nitrogen
stress
, on the other hand, has been estimated by quantifying the reflectivity of crop
canopy in images captured by a multispectral imaging sensor (Kim et al. 2001). This
sensor provides three simultaneous channels sensitive to the red, green, and infra-
red bands of the spectrum. Variations in leaf reflectivity correlate to shortages of
nitrogen because spectral reflectance is inversely correlated to the nitrogen content
of the crop. In a similar fashion, a
multispectral camera
also capturing three inde-
pendent channels sensitive to the red, green, and blue was set to detect and estimate
crop damage
caused by soybean rust severity (Cui et al. 2009). An indirect way to
predict yield in grapes is by quantifying the
vegetative vigor
of the plants. A simple
approach to acknowledge the vitality of vines is by comparing the amount of leaves
of the plants throughout the rows of a field in order to express it in a map; this rela-
tive comparison represents the spatial variability of plant development as a means
to anticipate production irregularities soon enough to introduce effective correc-
tions. The easiest way to enhance vigorous vegetal tissue in images is by mounting a
near-infrared (NIR) filter on the lens of a NIR-sensitive monochrome camera. This
straightforward procedure facilitates image processing routines and therefore alle-
viates the computational load of onboard processors. The sample images shown in
Figure 12.2 have been taken with a monochrome camera incorporating NIR filters.
The visual sensors used for 2-D surrounding awareness are
monocular cameras
,
that is, optical devices with one lens, which use a light-sensitive electronic array to
form the images. These sensitive arrays differ according to the band of the electro-
magnetic spectrum they are sensitive to. For navigation and obstacle detection, most
of the cameras incorporate arrays adapted to the visible spectrum (400-750 nm), often
extended to the NIR band (700-1400 nm). For precision agriculture applications, mul-
tiple stripes of the electromagnetic spectrum are used based on the specific property
being measured in the field. They normally range from the ultraviolet to the thermal
infrared. Multispectral cameras, in addition, are capable of capturing various images
simultaneously through the same lens by integrating different sensor arrays set up to
sense at a particular electromagnetic band, allowing a pixel-to-pixel correspondence.
After the perception sensor has been selected, the second decision to make relates
to its
position in the vehicle
. Similarly, it will also be determined by the application
pursued. Figure 12.2 depicts the three most common locations for a visual sensor
perceiving in the vicinity of an agricultural vehicle. The
top
(
zenithal
)
position
of
