(For USM Staff/Student Only)

EngLib USM > Ω School of Electrical & Electronic Engineering >

Feature extraction for butterfly classification using image processing

Feature extraction for butterfly classification using image processing / Yap Jin Hong
Kajian terhadap klasifikasi spesies rama-rama adalah sangat penting kepada entomologi kerana ia membantu dalam memahami dan mengesan habitat rama-rama. Kini, pengesanan dan penyetempatan rama-rama dilakukan secara manual oleh entomologi yang mahir serta sangat berpengalaman dengan mengenal setiap speses rama-rama melalui ciri morfologi pada rama-rama. Walau bagaimanapun, cara tradisional menangkap rama-rama secara menangkap dan melepas memerlukan kesabaran, membazirkan masa serta pengetahuan untuk mengenal pasti spesies rama-rama. Oleh sebab itu, satu system pintar yang boleh mengenal pasti spesies rama-rama secara automatik harus diwujudkan untuk menolong entomologi dalam penyelidikan tanpa mempunyai risiko untuk mencederakan rama-rama semasa menankap rama-rama menggunakan cara tradisional. Oleh itu, projeck yang dicadangkan ini bertujuan untuk membangunkan satu system klasifikasi rama-rama yang berasaskan pengekstrakan ciri dengan mengaplikasikan pemprosesan imej. Sistem ini dianggap akan mengenali species rama-rama dengan tepat menurut imej yang telah diproses. Imej yang ditangkap menggunakan alat-alat elektronik dan kamera akan diproces sebelum ia bersedia untuk menjalani pengenalan secara lanjut. Proses pengenalan rama-rama terdiri dari tiga bahagian, iaitu segmentasi imej, pengekstrakan ciri dan akhirnya, pengenalan. Tiga cara pengekstrakan ciri akan dibandingkan dalam kajian ini untuk mengetahui cara mana yang lebih sesuai untuk system ini. Antara cara pengekstrakan ialan, Law’s Texture Features (LTF), Gabor Filters (GF) dan Grey-Level Co-occurrence Matrix (GLCM). Manakala pengelas yang digunakan ialah K-Nearest Neighbours (kNN). 5-kali dan 10-kali ganda silang pengesahan teleh dilaksanakan untuk mencari nilai K dalan KNN dan menilai ketepatan ketiga-tiga kaedah analisis texture. Pangkalan data dikumpulkan dari Entopia, Penang Butterfly Farm dan melalui sumber-sumber dari internet. Sistem klasifikasi yang dicadangkan telah berjaya mengesan spesies rama-rama dan lebih banyak sampel spesis rama-rama boleh dikumpul melalui kaedah ini berbanding dengan kaedah konvensional. Oleh itu, kerja mengumpul data boleh dilakukan dengan lebih cekap dan cepat. _______________________________________________________________________________________________________ Study on the classification of butterfly species has become very vital to aids entomologist in tracking the habitats and to understand them. Currently, detection and localization of butterflies are done manually by entomologist which skilful and experience enough are capable in recognizing each species by their morphological characteristics of the butterfly. However, this catch-and-release method requires countless of time and patient as well as knowledge to accurately identifies the butterfly. Hence, there is a need for an intelligent system that can automatically identify the butterfly species and to assist entomologist in their research in the future without having the risks of harming the butterfly during the catch-and-release practise. Hence, this proposed project aims to develop a feature extraction based butterfly classification system that applying image processing which is expected to recognize butterfly species accurately according to the processed image. Image captured using electronic devices and cameras are then pre-processed before it is ready for further identification. As mentioned, the identification processes carried out consists of three parts which is the image segmentation, feature extraction and identification. Three types of texture feature extraction, the Law’s Texture Features (LTF), Gabor Filters (GF) and Grey-Level Co-occurrence Matrix (GLCM) are compared in this study to distinguish which methods are the most applicable to be used as feature extraction technique for this system. The classifier employed in this study is K-Nearest Neighbours (kNN) algorithm. 5-fold and 10-fold Cross Validation was carried to determine the value of K in kNN and evaluated the proposed texture analysis methods. The Databases are gathered from Entopia, Penang Butterfly Farm and through sources from the internet. The proposed butterfly species classification system has successfully detected the butterfly species. More samples of butterfly species can be collected via this method compare to conventional approach. Thus, the field work can be done more efficiently and less time consuming.
Contributor(s):
Yap Jin Hong - Author
Primary Item Type:
Final Year Project
Identifiers:
Accession Number : 875007224
Barcode : 00003107103
Language:
English
Subject Keywords:
butterfly species; entomologist; localization of butterflies
First presented to the public:
6/1/2017
Original Publication Date:
4/16/2018
Previously Published By:
Universiti Sains Malaysia
Place Of Publication:
School of Electrical & Electronic Engineering
Citation:
Extents:
Number of Pages - 129
License Grantor / Date Granted:
  / ( View License )
Date Deposited
2018-04-16 17:27:16.604
Date Last Updated
2019-01-07 11:24:32.9118
Submitter:
Mohd Jasnizam Mohd Salleh

All Versions

Thumbnail Name Version Created Date
Feature extraction for butterfly classification using image processing1 2018-04-16 17:27:16.604