Publication: Shape- and texture-based fish image recognition system
4
Issued Date
2013-11-12
Resource Type
ISSN
00755192
Other identifier(s)
2-s2.0-84887179880
Rights
Mahidol University
Rights Holder(s)
SCOPUS
Bibliographic Citation
Kasetsart Journal - Natural Science. Vol.47, No.4 (2013), 624-634
Suggested Citation
Chomtip Pornpanomchai, Benjamaporn Lurstwut, Pimprapai Leerasakultham, Waranat Kitiyanan Shape- and texture-based fish image recognition system. Kasetsart Journal - Natural Science. Vol.47, No.4 (2013), 624-634. Retrieved from: https://repository.li.mahidol.ac.th/handle/123456789/30951
Research Projects
Organizational Units
Authors
Journal Issue
Thesis
Title
Shape- and texture-based fish image recognition system
Other Contributor(s)
Abstract
This research developed a computer system capable of recognizing some fish images. The system known as the "shape- and texture-based fish image recognition system" (FIRS) consists of five subsystems-namely: 1) image acquisition, 2) image preprocessing 3) feature extraction, 4) image recognition and 5) result presentation. The experiment was conducted on 30 fish species, which consisted of 600 fish images as the training dataset and 300 fish images for testing. The system compared two recognition techniques-a Euclidean distance method (EDM) and artificial neural networks (ANN). The system was able to recognize all 30 species of the training fish images with a precision of 99.00 and 81.67% for the ANN and the EDM techniques, respectively. The average access times were 24.4 and 154.43 sec per image for the EDM and ANN techniques, respectively.