Agriculture Reference
In-Depth Information
agricultural environments without impairing the quality of the work in comparison
with that achieved by current manual methods. The cost of robotic systems must
be sufficiently low in order for them to be cost effective and additionally inherent
safe and reliable. Detailed mapping of fields for soil structure, texture, nutrients,
moisture content, pC and pH already allows precision automated and robotic hus-
bandry. In turn, this exploits genomic characteristics of cultivars allowing precise
predictions of sowing, planting, maturation and maturity, and harvesting. Intelligent
robotic technologies will cope competently with continuously changing crop mor-
phology and physiology and the environment. This demands information acquisi-
tion systems, including sensors, fusion algorithms and data analysis suitable for
field conditions.
The combined interest in precision agriculture, information technology, and au-
tonomous navigation has led to a growing interest in the generation of 3-dimensional
(3-D) maps of mobile equipment surroundings (Rovira-Mas
2008
). Creating three
dimensional terrain maps is achieved by combining the information captured with a
stereo camera, a localization sensor, and an inertial measurement unit, installed on
a mobile equipment platform. The perception engine comprises a compact stereo
camera that captures field scenes and generates 3D point clouds, which are trans-
formed to geodetic coordinates and assembled in a global field map. Results have
shown that stereo perception can provide the level of detail and accuracy needed
in the construction of 3D field maps for precision agriculture and field robotics ap-
plications. An alternative system replaces stereo perception with global satellite and
local positioning which may prove more accurate and reliable. In either case, the
automated application of fertilisers, agrochemicals and water is now becoming a
reality. The position of individual plants in rows and beds can be recorded into com-
puterised memory banks and utilised many times over for automated husbandry and
eventually harvesting operations. Protected crops can be produced by automated
seeding, propagation, grafting, transplanting, maintenance, harvesting, sorting and
packing. Already there are automated strawberry harvesters in Japan and rose har-
vesting and tomato de-leafing robots are in operation in the Netherlands. In China,
a wireless apple harvesting robot is emerging. Image analysis applied in ornamental
crops is being developed that will assesses leaf and flower cover, colour, uniformity,
and canopy height, and enable harvesting, grading, storage and dispatch. Post-har-
vest processing imagery will be able to detect visual defects by applying optical and
statistical methods identifying random colour and textures using multivariate image
and principal component analyses.
Image analysis is being developed for ornamental crops and the bedding plant
industries (Parsons et al.
2009
). Feed-forward artificial neural networks have been
used to segment top and side view images of three contrasting species of bedding
plants. The segmented images provide objective measurements of leaf and flower
cover, colour, uniformity and leaf canopy height. These systems may be applied
for crop grading at marketing or for monitoring and assessment of growing crops
within a glasshouse during all stages of production.
Glasshouse mechanization is being pioneered in The Netherlands (Henten
2006
).
Research demonstrates that the initial phases of plant production such as seeding,
