Agriculture Reference
In-Depth Information
occluded fruits using two-dimensional vision and execution rates between 14.6 and
19.7 s according to Lee and Slaughter (2004) are the two key weaknesses that need
to improve in order for these systems to be applied to mass harvesting machines.
Chinchuluun et al. (2006) built an automatic machine vision system for citrus
fruit yield estimation using charge-coupled device (CCD) cameras, ultrasonic sen-
sors, and a differential global positioning system (DGPS) to develop image process-
ing algorithm for fruit detection. Rather than measuring crops on platform beds or
transporting belts, this system detected fruits on trees before harvesting, and could
not only count fruits on trees but also estimate size. The coefficient of determination
(
R
2
) was found to be up to 0.83 for the number of fruit from manual counting versus
the number of fruit counted by the vision-based counting algorithm. However, the
coefficient between the number of fruit counted by the algorithm and actual har-
vested fruit was only 0.64.
On the other hand, efforts using weighing-based yield monitoring techniques
have been successful in several applications. Weighing-based techniques can be
classified by intermittent and continuous weighing methods. Measuring weight of a
truck loaded with fruits using weighing sensors is one common method of intermit-
tent weighing. In spite of its price competitiveness and high measuring accuracy,
it has been replaced by the continuous weighing system because of its time-saving
efficiency and the necessity to reestablish material flow (Dawson et al., 1976).
Whitney et al. (1999) integrated the geographical information system (GIS) and
GPS with a fruit road-siding truck in order to investigate weight-based yield map-
ping. Whitney et al. (1999) integrated a load cell weighing system at each corner
of the truck lift bed and pressure transducers on the lift cylinder in order to record
weight data for citrus. Each time the truck operator loaded a fruit tub, he would push
a button to enable the data logging system to record load cell and pressure transducer
readings, while at the same time acquiring positioning data from the GPS receiver.
Although their yield monitoring system was only a prototype, it successfully cre-
ated color-coded yield maps that were easily interpreted by growers. Whitney et al.
(2001) implemented a real-time kinematic (RTK) GPS with this yield monitor and
compared position accuracy with two other commercial DGPS systems.
Upadhyaya et al. (2006) developed an electronic weighing device with an impact
plate and a conveyor speed sensing system to measure mass flow of tomatoes.
Tomatoes impacted the plate as they dropped off the harvester boom conveyor,
where the impact force and conveyor speed data were recorded continuously on a
data logger. This weighing system was integrated into a commercial tomato har-
vester and tested during the 2004 and 2005 harvesting seasons. A weigh wagon was
used to verify the measurements of the impact type electric weighing system. The
results of tests suggested very good potential for impact type weighing systems with
coefficient of determination exceeding 0.96.
7.4 ROBOTIC TREE FRUIT HARVESTING
Robotic solutions for fresh market fruit and vegetable harvesting have been stud-
ied by numerous researchers around the world during the past several decades.
However, very few developments have become adopted and put into practice. The
