Agriculture Reference
In-Depth Information
are necessarily local and move with the vehicle. As a result, coordinates are valid
instantaneously but they lose their meaning as soon as the vehicle occupies a dif-
ferent location. Crop tracking for automatic guidance based on machine vision, for
example, benefits from this approach as images are discarded as soon as they are
processed, and the algorithm is usually programmed to estimate the instantaneous
relative position between the crop rows and the vehicle. However, precision agricul-
ture applications rely on the quantification of spatial variability within fields, which
can only be achieved with geographical references at a global scale. Fortunately,
all the advantages of local perception already described along this chapter can be
extended to global maps, which feature common origin and axes, and therefore liber-
ate coordinates from the motion of the vehicle. Satellite-based positioning systems,
such as GPS, provide the real-time global coordinates of vehicles, which facilitate
the composition of global maps by merging perception information gathered from
sequences of vehicle-referenced maps. The advantages of global maps are their sense
of completeness and their capacity to store historical data for better planning of
future tasks.
The move from multiple, local-based, vehicle-fixed, reference systems to a
global
frame
with a unique origin is not trivial. Apart from perception sensors (cameras,
lidars, or sonars) in charge of surrounding awareness, two fundamental types of
sensors need to be added to the intelligent system of the vehicle: a global position-
ing device and a sensor capable of estimating the vehicle attitude angles yaw, pitch,
and roll. Global positioning devices—only GPS is currently available—provide the
real-time localization of vehicles in geodetic coordinates: latitude, longitude, and
altitude. However, these coordinates are not convenient for the assemblage of global
maps in agricultural environments because their origin is remote, coordinates are
not Cartesian, and the calculation of distances requires spherical geometry. The
sphericity of the earth can be neglected for the majority of agricultural fields, and as
a result, an alternative coordinate system better fit to the situations encountered in
actual production sites is necessary. The Local Tangent Plane (LTP) system of coor-
dinates meets all the requirements desirable in agricultural situations: references
are global, the origin is set by the users at their convenience, the frame is Cartesian,
and LTP geometry is Euclidean. In addition, the horizontal plane contains the axes
east
and
north
, which are quite intuitive for field operators, with the third coordi-
nate being (
D
) the height measured perpendicular to the horizontal plane
E
-
N
. The
transformation from geodetic coordinates to LTP coordinates needs to be carried
out in real time as soon as GPS strings get to the onboard computer. This process
requires the selection of a reference ellipsoid—usually WGS84—and the applica-
tion of a sequence of mathematical expressions leading to the coordinates
E
-
N
-
D
(Rovira-Más et al. 2010). Given that vehicle-fixed coordinate frames move with the
vehicle, the entire frames are influenced by the attitude angles affecting the vehicle
in its motion. Because most of the fields are approximately flat, pitch and roll tend to
be insignificant and can be neglected in many cases. However, the yaw angle, also
known as the
vehicle heading
, is of great importance for the correct construction of
global maps. Farm equipment typically navigates following crop rows and tree lines,
but motion occurs in both (opposite) directions along the lanes and in perpendicular
direction around the headlands. In consequence, the orientation of the vehicle is
