Knowledge representation for durian varieties images using conceptual graph / Nur Syafikah Ibrahim

• Main Author: Ibrahim, Nur Syafikah
• Format: Thesis
• Language: English
• Published: 2016
• Online Access: https://ir.uitm.edu.my/id/eprint/17795/2/TM_NUR%20SYAFIKAH%20IBRAHIM%20CS%2016_5.pdf; https://ir.uitm.edu.my/id/eprint/17795/

Semantic Based Image Retrieval (SBIR) is an image retrieval approach mainly aims to improve the relevancy of the images retrieved. The researches in image retrieval were conducted in various domains and each domain requires specific queries. Knowledge Representation (KR) is a method under SBIR which represent the knowledge of a specific domain by using formal mathematical symbols. The existing hundreds of durian varieties which are currently registered in the Department of Agriculture Malaysia (DOA) make it a challenging task to differentiate the images of this crop. Hence, this research was intended to achieve three objectives. The first objective is to construct the Conceptual Graph (CG), which is one of the KR formalism to semantically represent the knowledge of durian varieties characteristics. The second objective is to employ the constructed CG in Knowledge Based Image Retrieval System (KBIRS). Meanwhile, the third objective is to evaluate the performance of the KBIRS. In this work, characteristics of 32 registered durian varieties were studied. There are three main characteristics that enable us to differentiate one variety from another variety which are fruit, aril (flesh) and spine (thorn) characteristics. These characteristics are called as concept types in CG. The KBIRS was tested by using 26 predefined queries and the retrieved results were evaluated by using the precision calculation. This precision result was then compared with the result in Exalead and Google Images search engine by using the same 26 predefined queries.