Pusat Pengajian Sains Matematik - Tesis
Browse
Recent Submissions
- PublicationImage Reconstruction By Dominant Point Detection And Least Squares Rational Quadratic BéZier(2019-02)Lim, Ai HuiThe techniques of boundary-based dominant point detection and curve fitting are important in the analysis of digital images. They provide key information of images for further processing and analyzing in lower amount of data, hence speed up the pro- cess of analysis. This relates to the processes of shape description, pattern recognition and image understanding. Boundary data of binary images are extracted and then rep- resented in coordinates of planar form. These data are then used for the processes of support region determination, dominant point detection and lastly, curve fitting. Re- gion of support plays a vital role in the process of locating precise dominant points of a set of boundary data. Support region of a point is a subset of boundary data points which bear the maximum level of linearity. Techniques consist of geometrical and sta- tistical approaches are presented to find proper support region. Two existing methods in finding support region are improved to enhance their performance. The methods are named as unit bending value and Marji & Siy method. Besides, three new methods are proposed in this work to determine support region, namely eigen-ratio, eccentric- ity with eigenvalues and convex hull area-length ratio. Curvature measurement is a common method to extract dominant points. Geometrical and statistical methods are applied to estimate curvatures and hence locating the dominant points. Data points with high curvature will be selected as dominant points to represent the shape of the xix
- PublicationA Study Of Statistical Methods In Aids Modelling(1997-01)Ong, Hong ChoonThe objective of this research is to do a critical survey on some of the statistical methods that have been developed on the modelling of the hiv/aids epidemic and to study the development of various methods which have been used from the year 1985 until the year 1995 to better understand the aids epidemic. A study of how the various statistical issues related to the collection and the nature of the hiv/aids data is made. These data are adjusted before they are incorporated into the model. The statistical issues include biases in cohort prevalence, censoring of the data, delay in reporting, changes in aids surveillance definition and many others.
- PublicationTopological Data Analysis In Analyzing The Similarity Of Air Pollutants’ Behavior Across Air Quality Monitoring Stations In Malaysia(2025-07)Nwabuisi, Madukpe VineAir pollution is increasingly recognized as a major global issue resulting from industrial processes, urban development, transportation emissions, and a wide range of human activities. This study investigates air pollutant behavior through a qualitative approach, utilizing topological data analysis (TDA) techniques, the conventional Mapper (CM), and its variant Ball Mapper (BM). It analyzes six major air pollutants (NO2, PM10, PM2.5, O3, CO, and SO2) collected from 60 air quality monitoring stations across Malaysia. The topological graphs generated using the CM algorithm revealed clusters of regions and stations exhibiting similar pollutant behavior. The results showed similar patterns in PM10 behavior across the years 2013, 2015, and 2020.
- PublicationTrajectory Prediction And Near Miss Detection Using Social Distancing Monitoring And Distance Based Proximal Indicators(2025-07)Lim, Lek MingDespite efforts to improve road safety, accidents persist due to insufficient evidence from manual police reporting, non-optimized detection algorithms, and technical limitations in real-time video processing and modelling. This study focuses on detecting and tracking vehicles within a monitoring system and analysing near-miss incidents (black spot and blind spot), specifically examining the influence of video quality on detection performance using advanced model detectors (YOLOv3-tiny, YOLOv4-tiny, YOLOv5, YOLOv7, YOLOv7+G3HN, YOLOv7+CNeB, and Faster RCNN). The experiment employed methods for vehicle detection and trajectory prediction through the monitoring system.
- PublicationJ-Type Random 2,3 Satisfiability Reverse Analysis Topological-Based Method In Discrete Hopfield Neural Network(2025-07)Jiang, XiaofengThe development of satisfiability logical representation in Discrete Hopfield NeuralNetworks has evolved into a more flexible structure, allowing for both systematic and non-systematic logic. However, the main issue with the exiting flexible satisfiability representation is the appearance of first order clauses which will degrade the quality of the logical rule as a neuron representation and leads to overfitting issue. Therefore, this thesis proposes a hybrid higher order satisfiability logical rule named J-Type Random 2,3 Satisfiability. The proposed logical rule randomly generates structures based on either second order, third order, or a combination of both clauses. The behavior of the proposed logical rule will be evaluated using various performance metrics in terms of training error, retrieval error, energy analysis, and similarity analysis. Simulation results indicate that the proposed model achieves the highest global minimum ratio with an average value of 0.4674. The newly proposed model will be incorporated into the logic mining model known as topology based J-type random 2,3 satisfiability reverse analysis. The unsupervised attribute selection method called topological data analysis is employed in selecting most significant attributes. Meanwhile, permutation operation improves the ability of the logic mining model to retrieve the best induced logic that can extract patterns from the dataset. The proposed logic mining model demonstrates superior performance compared with existing state-of-the-art logic mining models with an average accuracy of 0.8375 in the selected various real-life datasets.