Agriculture Reference
In-Depth Information
with color CCD cameras stationed at six different angles to provide all side fruit
images. Lighting is provided through halogen lamps or light-emitting diodes (LEDs)
with polarizing filters that are attached to camera lenses to eliminate halation on
glossy surfaces of fruits. In the system, an x-ray imaging device and a NIR inspector
are also installed online to obtain internal qualities.
When the cameras acquire images, the bottom image of fruit is not easy to obtain.
The specially designed roller pins turn over the fruit at 180
, ensuring that a full
view is acquired by the cameras from the top. Basically two inspection stages of the
system can be identified: external fruit inspection and internal fruit inspection stage.
In the external inspection stage, images from the CCD cameras set under random
trigger mode are copied to the image grabber board fitted on the image processing
computer whenever a trigger occurs. The images are processed using specific algo-
rithms for detecting image features of color, size, bruises, and shape. From the fruit
images, the following features are commonly extracted for fruit inspection: (1) size
(maximum and minimum diameter, area, and extrapolated diameter), (2) color (color
space based on HSI values, chromaticity, and
L*a* b*
), (3) shape (ratio between max-
imum length and width, complexity, circularity factor, distance from gravity center
to fruit border), and (4) defect (binary results based on color difference on R-G
derived images and on-edge detection operation).
For internal fruit inspection, an NIR spectroscopy determines the sugar content
(brix equivalent) and acidity level of the fruits from the light wavelengths received
after light is transmitted through the fruit. In addition, the NIR inspector measures
the granulation level of the fruit, which indicates the inside water content of orange
fruit. Rind puffing, a biological defect that occurs in oranges, is inspected using the
x-ray sensor. Output signals from PCs collecting sensing data are transmitted to the
judgment computer, where the final grading decision is made based on fruit appear-
ance features and internal quality measurements.
An important feature of the system design is that it is adaptable to the inspec-
tion of many other products such as potato, tomato, persimmon, and kiwi fruit with
adjustments changing software by the process cords. Several lines for orange fruit
inspection combined with high conveyance and high-speed computers enable the
system to handle large batches of fruit product at high speeds. All the information
from receipt of fruit at the collection site to grading, packing, and shipping is stored
in a server for each producer for future reference (Njoroge et al., 2002).
°
14.3.3.2 Grading Robot
Peaches, pears, and apples are not suitable for the above-described conveyors and
handling systems because they are easily damaged on the systems. A grading robot
system, which automatically provides fruit from containers and inspects all sides of
the fruit, was developed (Kondo, 2003). It has two Cartesian coordinate robots called
the providing robot and the grading robot. The grading robot consists of a 3-DOF
manipulator, 12 suction pads as end effectors, 12 color TV cameras, and 28 lighting
devices with polarizing filters, whereas the providing robot has similar manipula-
tor and end effectors, but has no machine vision. Twelve fruits are sucked up by a
manipulator at a time and 12 bottom images of fruits are acquired during the time
the manipulator is moving to carriers on a conveyor line. Before releasing the fruits
