(Agricultural Information
Research, 18(1), 2009, 1-7)
Non-destructive Detection of Damaged Raw Unhulled
Grains Using Image Processing
--Detection of Damaged Grains by
Stinkbugs--
Yosuke Kubota, Kazuhiro
Nakano, Satoshi Yumoto and Yasuhiro Higuchi
Summary
The purpose of
this study is to detect damaged rice grains, which are unhulled and internally
damaged by stinkbugs, using an image-processing system. The experimental
apparatus mainly consists of a multispectral camera, fiber-optic light sources,
plane-emission light sources, sample plate and personal computer. More
specifically, the transmitted light images of 50 grains on the sample plate were
generated when the grains were illuminated from the back of the sample plate.
Furthermore, while the sample plate was fixed, the light sources were changed
over to the fiber-optic light sources irradiating the grains equally from four
corners on the upper side of the sample plate. In succession, the raw grains
were hulled and visually confirmed to see if any damage had been caused by
stinkbugs. The images were processed in the methods using binary-conversion
processing, contraction processing and mask operation. In this regard, the mean
gray level of each normal grain and the number of pixels within the damaged
parts of grains were obtained to draw a scatter chart from which the linear
discriminant function was derived. As a result, the undamaged grains and
stinkbug-damaged grains were discriminated from each other. It was found that
the total discrimination rate was 94.6 percent. In conclusion, the possibility
of detecting stinkbug-damaged grains using raw unhulled grains can be suggested
when the image-processing algorithm constructed during this research comes into
use.