| Summary: | 碩士 === 國立成功大學 === 造船工程學系研究所 === 85 === traditionally , the gradif squid on the squid fishing boat is totally l
absor-intensive. Usually, it takes nearly 10 hours for a working group to co
mplete the sorting . There are 10 persons around in a working group. It is n
oteconomical and efficient. Therefore, in this paper , we apply the theory
of Neral network and the technique of digital image processing on the gr
ading . The processing of the grading is summarized as follows. 1. We
set a suitable threshold and use a median filter to get a binary imag
e from the initial imiage of the squid. 2. From the binary image, we get fe
atures of the squid,like the area , width , the vertical and horizontal
length of the squid. 3. We set three standard training model A , B
and C accord- ing to the features we get above , and input the featu-
res to the Backpropagation Network for Training. 4. Compare the output
of the Neural network with the true squid , we get the best mode A4 .
5. We apply the A4 model on the image processing of sorting, machine t
o complete the grading of the squid.
|
|---|
