Identification of cultivated rice MR 263 seed and weedy rice seed variants using CCD camera based-machine vision system

The main purpose of this study was to develop rice seed identification research prototype system to classify cultivated rice and weedy rice seeds variants using machine vision system through the extraction of morphological, colour, and textural features of the seeds. Five different types of weedy r...

Full description

Saved in:
Bibliographic Details
Main Author: Aimi Athirah, Aznan
Format: Thesis
Language:English
Subjects:
Online Access:http://dspace.unimap.edu.my:80/xmlui/bitstream/123456789/61514/1/Page%201-24.pdf
http://dspace.unimap.edu.my:80/xmlui/bitstream/123456789/61514/2/Full%20text.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
id my-unimap-61514
record_format uketd_dc
spelling my-unimap-615142023-04-04T02:12:24Z Identification of cultivated rice MR 263 seed and weedy rice seed variants using CCD camera based-machine vision system Aimi Athirah, Aznan Professor Ir. Dr. Ibni Hajar Bin Hj. Rukunudin The main purpose of this study was to develop rice seed identification research prototype system to classify cultivated rice and weedy rice seeds variants using machine vision system through the extraction of morphological, colour, and textural features of the seeds. Five different types of weedy rice seeds variants samples of open panicle, close panicle and awn type were collected from several commercial farms in Kedah. The MR 263 seed was obtained from a commercial rice seed bag from a local supplier. In this study, seed samples were consisted of 600 seeds of MR 263 and 600 seeds from weedy rice seed variants group. Images of the rice seed samples were acquired using a charge coupled device (CCD) colour camera. Laboratory Virtual Instrument Engineering Workbench (LabVIEW) development environment was used to program the image processing, features extraction and the classification analysis. There was 12 morphological, 6 colour and 5 textural features were extracted from the seed images. Four types of classification model namely morphology, colour, texture and morphologycolour- texture models were established based on the extracted data. Each of the models was analyzed for feature selection using stepwise discriminant analysis (SDA) to develop the optimized features model. Then, the original and optimized features models were analyzed using 3 classifiers; discriminant function analysis (DFA), support vector machine (SVM) and neural network (NN). Analysis of variance (ANOVA) was conducted on the 3 classifiers to evaluate the mean classification accuracy levels of the 8 extracted features models developed. The ANOVA showed that there is no significant difference of mean classification accuracies between the 3 classifiers. The classification results using morphology-colour-texture features model was found to obtain higher classification accuracy levels as compared to the single feature models. An identification system was developed in the LabVIEW to classify the cultivated rice MR 263 and weedy rice seed groups using optimized features of the morphology-colourtexture model in DFA. The developed system was able to classify both seed groups at 99.4% accuracy level using testing data set. Universiti Malaysia Perlis (UniMAP) 2015 Thesis en http://dspace.unimap.edu.my:80/xmlui/handle/123456789/61514 http://dspace.unimap.edu.my:80/xmlui/bitstream/123456789/61514/1/Page%201-24.pdf d1dca3f1e125974a5e3b5e9e025f5b43 http://dspace.unimap.edu.my:80/xmlui/bitstream/123456789/61514/2/Full%20text.pdf 0df4e88ff5a326c92b18a24a4de67f6f http://dspace.unimap.edu.my:80/xmlui/bitstream/123456789/61514/3/license.txt 8a4605be74aa9ea9d79846c1fba20a33 Rice seed Rice seed industry -- Malaysia MR 263 -- Rice variety Paddy production industry -- Malaysia School of Bioprocess Engineering
institution Universiti Malaysia Perlis
collection UniMAP Institutional Repository
language English
advisor Professor Ir. Dr. Ibni Hajar Bin Hj. Rukunudin
topic Rice seed
Rice seed industry -- Malaysia
MR 263 -- Rice variety
Paddy production industry -- Malaysia
spellingShingle Rice seed
Rice seed industry -- Malaysia
MR 263 -- Rice variety
Paddy production industry -- Malaysia
Aimi Athirah, Aznan
Identification of cultivated rice MR 263 seed and weedy rice seed variants using CCD camera based-machine vision system
description The main purpose of this study was to develop rice seed identification research prototype system to classify cultivated rice and weedy rice seeds variants using machine vision system through the extraction of morphological, colour, and textural features of the seeds. Five different types of weedy rice seeds variants samples of open panicle, close panicle and awn type were collected from several commercial farms in Kedah. The MR 263 seed was obtained from a commercial rice seed bag from a local supplier. In this study, seed samples were consisted of 600 seeds of MR 263 and 600 seeds from weedy rice seed variants group. Images of the rice seed samples were acquired using a charge coupled device (CCD) colour camera. Laboratory Virtual Instrument Engineering Workbench (LabVIEW) development environment was used to program the image processing, features extraction and the classification analysis. There was 12 morphological, 6 colour and 5 textural features were extracted from the seed images. Four types of classification model namely morphology, colour, texture and morphologycolour- texture models were established based on the extracted data. Each of the models was analyzed for feature selection using stepwise discriminant analysis (SDA) to develop the optimized features model. Then, the original and optimized features models were analyzed using 3 classifiers; discriminant function analysis (DFA), support vector machine (SVM) and neural network (NN). Analysis of variance (ANOVA) was conducted on the 3 classifiers to evaluate the mean classification accuracy levels of the 8 extracted features models developed. The ANOVA showed that there is no significant difference of mean classification accuracies between the 3 classifiers. The classification results using morphology-colour-texture features model was found to obtain higher classification accuracy levels as compared to the single feature models. An identification system was developed in the LabVIEW to classify the cultivated rice MR 263 and weedy rice seed groups using optimized features of the morphology-colourtexture model in DFA. The developed system was able to classify both seed groups at 99.4% accuracy level using testing data set.
format Thesis
author Aimi Athirah, Aznan
author_facet Aimi Athirah, Aznan
author_sort Aimi Athirah, Aznan
title Identification of cultivated rice MR 263 seed and weedy rice seed variants using CCD camera based-machine vision system
title_short Identification of cultivated rice MR 263 seed and weedy rice seed variants using CCD camera based-machine vision system
title_full Identification of cultivated rice MR 263 seed and weedy rice seed variants using CCD camera based-machine vision system
title_fullStr Identification of cultivated rice MR 263 seed and weedy rice seed variants using CCD camera based-machine vision system
title_full_unstemmed Identification of cultivated rice MR 263 seed and weedy rice seed variants using CCD camera based-machine vision system
title_sort identification of cultivated rice mr 263 seed and weedy rice seed variants using ccd camera based-machine vision system
granting_institution Universiti Malaysia Perlis (UniMAP)
granting_department School of Bioprocess Engineering
url http://dspace.unimap.edu.my:80/xmlui/bitstream/123456789/61514/1/Page%201-24.pdf
http://dspace.unimap.edu.my:80/xmlui/bitstream/123456789/61514/2/Full%20text.pdf
_version_ 1776104220605284352