Image Recognition of Rotifers with Machine Vision
博士 === 國立臺灣大學 === 生物產業機電工程學研究所 === 89 === A machine vision system has been developed to substitute for human examination in the identification and classification, by shape analysis, of rotifers under the microscope. This automated system, consisting of both motion control and pattern recognition uni...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | en_US |
Published: |
2001
|
Online Access: | http://ndltd.ncl.edu.tw/handle/61481091941487402282 |
id |
ndltd-TW-089NTU00415023 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-089NTU004150232016-07-04T04:17:05Z http://ndltd.ncl.edu.tw/handle/61481091941487402282 Image Recognition of Rotifers with Machine Vision 機器視覺應用於輪蟲影像之辦識 Chan-Yun Yang 楊棧雲 博士 國立臺灣大學 生物產業機電工程學研究所 89 A machine vision system has been developed to substitute for human examination in the identification and classification, by shape analysis, of rotifers under the microscope. This automated system, consisting of both motion control and pattern recognition units, was designed to operate on a highly precise platform. An industrial personal computer was then used to control the operation platform, which was comprised of a two-dimensional, linear pulse motor and a microscope. The computer’s high processing speed also enabled it to perform the tasks of image processing and model discrimination. An algorithm thought suitable for image processing was proposed and then, after scrutiny, the results were discussed in detail. In this study, rotifers were classified into clearly defined types despite the presence of debris in the form of both scum in the stagnant water and dead rotifers. To this end, a two-phase discrimination model based on shape analysis was devised; the first stage involved separating the debris from the rotifers, while the second involved placing rotifers into one of three groups. A set of shape descriptors, including geometric and moment features, was then extracted from the images. These descriptors were required to display a prerequisite degree of RST (Rotation, Scaling, and Translation) invariance. Shape analysis was proved to be a particularly effective approach to classification as it was found to be approximately 92% accurate, on average. The results gained by applying the various classification approaches, including both the neural and statistical methods, were compared. Several modified models, intended to maximize classification precision, were suggested and later shown to yield improvement in either the first or second stage. The machine vision system with shape analysis and the two-staged model of discrimination both significantly reduced the need for human labor in the classification of rotifers. Jui-Jen Chou 周瑞仁 2001 學位論文 ; thesis 168 en_US |
collection |
NDLTD |
language |
en_US |
format |
Others
|
sources |
NDLTD |
description |
博士 === 國立臺灣大學 === 生物產業機電工程學研究所 === 89 === A machine vision system has been developed to substitute for human examination in the identification and classification, by shape analysis, of rotifers under the microscope. This automated system, consisting of both motion control and pattern recognition units, was designed to operate on a highly precise platform. An industrial personal computer was then used to control the operation platform, which was comprised of a two-dimensional, linear pulse motor and a microscope. The computer’s high processing speed also enabled it to perform the tasks of image processing and model discrimination.
An algorithm thought suitable for image processing was proposed and then, after scrutiny, the results were discussed in detail. In this study, rotifers were classified into clearly defined types despite the presence of debris in the form of both scum in the stagnant water and dead rotifers. To this end, a two-phase discrimination model based on shape analysis was devised; the first stage involved separating the debris from the rotifers, while the second involved placing rotifers into one of three groups. A set of shape descriptors, including geometric and moment features, was then extracted from the images. These descriptors were required to display a prerequisite degree of RST (Rotation, Scaling, and Translation) invariance.
Shape analysis was proved to be a particularly effective approach to classification as it was found to be approximately 92% accurate, on average. The results gained by applying the various classification approaches, including both the neural and statistical methods, were compared. Several modified models, intended to maximize classification precision, were suggested and later shown to yield improvement in either the first or second stage. The machine vision system with shape analysis and the two-staged model of discrimination both significantly reduced the need for human labor in the classification of rotifers.
|
author2 |
Jui-Jen Chou |
author_facet |
Jui-Jen Chou Chan-Yun Yang 楊棧雲 |
author |
Chan-Yun Yang 楊棧雲 |
spellingShingle |
Chan-Yun Yang 楊棧雲 Image Recognition of Rotifers with Machine Vision |
author_sort |
Chan-Yun Yang |
title |
Image Recognition of Rotifers with Machine Vision |
title_short |
Image Recognition of Rotifers with Machine Vision |
title_full |
Image Recognition of Rotifers with Machine Vision |
title_fullStr |
Image Recognition of Rotifers with Machine Vision |
title_full_unstemmed |
Image Recognition of Rotifers with Machine Vision |
title_sort |
image recognition of rotifers with machine vision |
publishDate |
2001 |
url |
http://ndltd.ncl.edu.tw/handle/61481091941487402282 |
work_keys_str_mv |
AT chanyunyang imagerecognitionofrotiferswithmachinevision AT yángzhànyún imagerecognitionofrotiferswithmachinevision AT chanyunyang jīqìshìjuéyīngyòngyúlúnchóngyǐngxiàngzhībànshí AT yángzhànyún jīqìshìjuéyīngyòngyúlúnchóngyǐngxiàngzhībànshí |
_version_ |
1718334194154733568 |