Date of Award

2006

Degree Type

Thesis

Degree Name

Bachelor of Science Honours

School

School of Computer and Information Science

Faculty

Faculty of Computing, Health and Science

First Advisor

Dr Jitian Xiao

Second Advisor

Michael Collins

Abstract

One of the most expensive operations in a spatial database is spatial join processing. This study focuses on how to improve the performance of such processing. The main objective is to reduce the Input/Output (I/O) cost of the spatial join process by using a technique called cluster-scheduling. Generally, the spatial join is processed in two steps, namely filtering and refinement. The cluster-scheduling technique is performed after the filtering step and before the refinement step and is part of the housekeeping phase. The key point of this technique is to realise order wherein two consecutive clusters in the sequence have maximal overlapping objects. However, finding the maximal overlapping order has been shown to be Nondeterministic Polynomial-time (NP)-complete. This study proposes an algorithm to provide approximate maximal overlapping (AMO) order in a Cluster Overlapping (CO) graph. The study proposes the use of an efficient maximum weighted matching algorithm to solve the problem of finding AMO order. As a result, the I/O cost in spatial join processing can be minimised.

Share

 
COinS