Saad, Puteh (2003) Trademark image classification approaches using neural network and rough set theory. PhD thesis, Universiti Teknologi Malaysia, Faculty of Computer Science and Information System.
Restricted to Repository staff only
PDF (Table of Contents)
PDF (1st Chapter)
The critical step in automatic trademark matching is to extract trademark features from the database automatically and reliably. However, the performance of existing algorithms rely heavily on the size of the database. It is essential to incorporate an eficient classification technique to partition the database in order to ensure the performance of an automatic trademark matching system is robust with respect to the increase in the database size. Two new approaches are proposed to classify trademark images. The approaches contain five major stages, namely: image acquisition, image preprocessing, feature extraction, data transformation and classification. Feature normalization and data discretization techniques are utilized to perform the data transformation phase. An Adaptive Multi Layer Perceptron (MLP) embedded with an enhanced Backpropagation (BP) algorithm and Rough Set Theory are applied to classify the images. Experimental results reveal that the Adaptive MLP embedded with the enhanced BP algorithm exhibits a faster convergence rate than the classical BP algorithm. In conclusion, the Adaptive MLP outperforms Rough Set Theory in terms of speed, accuracy and sample size.
|Item Type:||Thesis (PhD)|
|Additional Information:||Thesis (Doctor of Philosophy - Computer science) - Universiti Teknologi Malaysia, 2004; Supervisor : Prof. Dr Safaai Deris|
|Uncontrolled Keywords:||boolean reasoning, learning, supervised classification, discretization, reasoning, boolean logic, neural networks|
|Subjects:||Q Science > QA Mathematics > QA75 Electronic computers. Computer science|
|Divisions:||Computer Science and Information System|
|Deposited By:||Ms Zalinda Shuratman|
|Deposited On:||17 Nov 2008 08:14|
|Last Modified:||03 Sep 2012 05:37|
Repository Staff Only: item control page