Enhanced And Automated Approaches For Fish Recognition And Classification System
Loading...
Date
2011-06
Authors
Samma, Ali Salem Ali
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
Recognition and classification of fish images with high degree of accuracy and efficiency
can be a difficult task due to fish being very similar to the background, missing of some features
and high cost of computation. The aim of this thesis is to overcome these problems by proposing
methods that overcome the problem of missing features of fish. The problems of high cost
of computation, inaccurate extraction and representation of features of fish, and inaccurate and
inefficient selection of desirable shape for fish recognition and classification are also addressed.
Furthermore, for automated detection and extraction of the fish, K-means and background subtraction
approaches for image segmentation are enhanced. An enhanced approach for shape
representation that combines run-length method and modified chain code method with region
information is also proposed. An enhanced approach for shape description that uses slope is
also proposed to reduce the computation time. For more accurate and efficient detection of
the critical points for a shape, a technique that combines skeleton and boundary information is
proposed. Finally, for a more accurate classification of fish, methods that use principal component
analysis and genetic algorithm with two methods of the support vector machine are
proposed. By utilising the above-mention enhanced approaches, an automated fish recognition
and classification system is also established. The enhanced methods make the fish recognition
and classification system achieves 98.4% accuracy and more accurate than the existing systems
as shown by the evaluation that has been carried out.
Description
Keywords
Fish Recognition , Classification System