Agriculture Reference
In-Depth Information
camera and capture board or PC), and image acquisition software and processing for
feature extraction and understanding. In fruit grading facilities or postharvest operations,
it is necessary to handle various type products in different seasons, because their proper-
ties depend not only on varieties but also on growing environmental conditions.
The most important factor is the lighting device when the machine vision system is
constructed. In fruit grading operations, in particular, many fruits have glossy surfaces
because of the cuticular layers on surfaces. This may sometimes cause halation on
the surface, where pixel information is lost, which makes it difficult to acquire high-
quality images. It is known that irregular fruit shapes often make unexpected halation
occur at local places, or the unevenness of illumination may be found on the fruit
surfaces, even when illumination conditions including camera settings are perfectly
adjusted for the fruit variety. The number of halation sometimes exceeds the number
of light sources because of the fruit's irregular shape. In addition to this problem, sur-
rounding devices and walls are reflected on the very glossy surface (just like a mirror).
Even when a dome where secondary or tertiary reflected light softly radiated the glossy
fruit was used for image acquisition, the dome walls or/and ceiling are still reflected in
the fruit surface. Elimination of halation and the surrounding's reflection is the most
important issue when constructing a machine vision system as well as to ensure a uni-
form illumination condition for fruit inspection. Kondo (2006) showed a method to use
a polarizing (PL) filter in front of the lighting device.
Figure 14.7(a) and (b) shows comparison of PL filtering image and non-PL filter-
ing image of a green apple. Although Figure 14.7(a) has no halation, four points are
observed on the apple surface in Figure 14.7(b). Two PL filters were used in front of
the lighting device and the camera lens, which were adjusted to eliminate the hala-
tion on output image from the camera. As for lighting device, halogen lamps were
frequently used so far, because of their higher brightness, higher color rendering,
cheaper price, and long shelf life. Recently, however, LEDs are getting more and
more popular because of the easy arrangement of the lighting device's shape and
color components including NIR and ultraviolet regions, higher response, and very
long product life that requires no maintenance for a long time. Monochrome cameras
show only the brightness of objects. Its intensity can indicate product surface gloss
such as eggplant fruits (Kondo et al., 2007), but it can also measure product size
and shape. Figure 14.7(c) and (d) shows images of different quality eggplant fruits. In
Figure 14.7(c) and (d), halation was intentionally made on fruit surfaces by three line
light sources so that internal quality could be predicted, because glossy surface egg-
plants are soft flesh, whereas dull surface eggplants are firm or old. From Figure 14.7
(c) and (d), it is observed that the left eggplant has much halation and is predicted to be
soft and fresh. The right one has a partially dull surface especially near the blossom
end. Sensitivities of monochrome CCDs usually range from visible to infrared regions
(many current CCDs have 400-1000 nm sensitivity). Because agricultural product
reflectance in the infrared region (700-1100 nm) is higher than that in the visible
region (400-700 nm), it can be said that such CCD cameras have the advantage of
being able to distinguish agricultural products from other objects or the background.
UV-A light sometimes excites fluorescent substances in agricultural products. It
is well known that rotten or injured orange fruit skins fluoresce by 365 nm UV light
(Uozumi, 1987) so that fluorescent images can be used for detecting damaged fruits
