The researchers in Computer Science have a wide range of potential projects in their respective areas.  This page provides a list of people and possible projects that are on offer.  If you are interested in a specific project or an specific area, feel free to contact the person directly.

• Dr Luigi Barone
• W/Prof Mohammed Bennamoun
• Prof Rachel Cardell-Oliver
• Prof Amitava Datta
• Dr Rowan Davies
• Asst/Prof Tim French
• Assoc/Prof Bruce Gardiner
• Assoc/Prof Eun-Jung Holden
• Assoc/Prof Du Huynh
• Dr Joel Kelso
• Asst/Prof Wei Liu
• Assoc/Prof Cara MacNish
• Assoc/Prof Chris McDonald
• W/Prof George Milne
• Dr Peter Pivonka
• Prof Mark Reynolds
• Prof Tim Sercomb
• Dr Ferdous Sohel
• Assoc/Prof Nick Spadaccini
• Assoc/Prof Lyndon While
• Prof Michael Wise
• 
  
• International Centre for Radio Astronomy Research Projects

Dr Luigi Barone

Dr. Luigi Barone has been working with evolutionary algorithms (EAs) for over ten years and has a detailed knowledge of the different kinds of these algorithms, as well as hands-on experience in implementing many of them. His research has been both into practical applications of EAs as well as theoretical implications arising from their use. Dr. Barone has experience in applying these types of optimisation techniques to many different kinds of problems including adaptive learning, opponent modelling, scheduling, and engineering design problems. Dr. Barone has won several competitive research grants for funding work in this area.

Projects

• Investigating the effects of noise on evolutionary algorithms
• Investigating heuristics for improving the performance of hypervolume calculation algorithms
• Distributed agent control for WAMBOT
• Adaptive intelligent play for the game of Pacman using evolutionary algorithms

Experience has shown that it can be very beneficial for research students to have a group of people with related interests to share ideas with. A student undertaking any of the above projects is expected to join the Adaptive Systems Research Group and will be expected to attend and contribute to group meetings and discussions. Such a student will be housed in the Adaptive Systems Research Group Laboratory in G.11 of the Computer Science building.

W/Prof Mohammed Bennamoun

• Image processing
• Computer vision (3D object recognition)
• Robotics

• Hyper-spectral Imaging for Early Detection of Human Skin Precancerous lesions

  Melanoma is the most severe form of skin cancers. This project aims to develop a new method for early detection of pre-malignant lesions in human skin using hyperspectral imaging. Hyperspectral imaging provides a visual image of the skin-lesion and a set of spectral images acquired at different wavelengths that can be used for characterizing the chemical spectra of any point in that image. Powerful image processing, multivariate statistical analysis and pattern recognition methods will be customized to uniquely characterize these spectra in order to provide a cost-effective screening technology for early detecting precancerous skin-lesions which will play a vital role in saving lives.

  A Hyper-spectral camera can provides wealthy but complex information about the chemical constituents of biological samples that can be used in search of precancerous lesions. Essentially, a hyper-spectral camera collects both spatial and spectral information from a scene of interest, builds a hyper-spectral data-cube from these image slices, and presents a highly resolved, rendered image to the investigator. Hyper-spectral imaging provides a researcher or clinician with a visual image of the biological sample of interest, and a set of spectral images acquired at different wavelength that can be used for characterizing the chemical spectra of any point or location in that image. Coupled with a hyper-spectral imaging system Raman spectroscopy of human tissues can be used as a very sensitive, non-invasive and non-subjective tool for the detection and localization of tumoral nests.

• Audio-Visual Speech and Speaker Recognition Joint project with Assoc/Prof Roberto Togneri of EE

Prof Rachel Cardell-Oliver

• Measuring the Programming Process

  STREAM is a programming process developed by Caspersen and Kolling for novice programmers. A STREAM-like process is used in UWA's Java and Software Engineering first year units. We have developed a metrics framework based on the STREAM model for measuring the quality of programs. So far, this has been used to analyse submitted student programs.  But an even more interesting problem is to measure the quality of programs produced during the programming process as students develop code. This project aims to measure novices programs during the development process. The long term aim is to improve understanding of the software development process.  It may also be possible to apply these techniques in professional software engineering settings.

• Automatic Detection of Code and Design Smells

  Code and design smells are poor solutions to recurring implementation and design problems. They may hinder the evolution of a system by making it hard for software engineers to carry out changes. Moha et al have proposed three contributions to the research field related to code and design smells: 1) Decor, a method that embodies and defines all the steps necessary for the specification and detection of code and design smells, 2) Detex, a detection technique that instantiates this method, and 3) an empirical validation in terms of precision and recall of Detex.

  This project is to apply the DECOR method to detect code smells for a software corpus of your choice - either student projects or professional open source software.

  Background: DECOR: A Method for the Specification and Detection of Code and Design Smells, Moha et al, Transations on Software Engineering, 36(1) 2010

• Testing Graphical User Interfaces

  Testing and learning to correctly implement a Graphical User Interface (GUIs) program is difficult for several reasons. Students often encounter problems with error handling, navigation, finding bugs and using the language libraries.  This project is to develop tools to support students to improve the quality of their Java Swing GUI programs. We propose a system based on the Java Abbot API for semi-automated support that contains a mixture of development tools, test cases, guided tutorial exercises and checklists. The system will be evaluated by analysing the quality of student GUI programming assignments developed both with and without the tool support.
  • Rieky Barady's 2009 Honours thesis
  • Abbot API

• Sensor Networks for Australian Sheep Monitoring

  Scientists from the University of Western Australia’s School of Animal Biology are looking for a way to monitor the movement of a flocks of 10,000 to 15,000 sheep in Western Australia’s rangeland near Meekathara. The main problem is to remotely monitor and possibly control the movement of flocks of sheep, answering such questions as “where are the sheep?”, “are any sheep separated from the main flock?”, “are the sheep near water?” A possible solution to this problem is to develop a sensor network of low costs nodes, each mounted on one sheep (but not the whole flock), combined with some higher power nodes on selected sheep, or high power nodes at fixed locations.  The network must be able to communicate the current status of a sheep flock over a distance of 50 to 100 kms a farm base station 4 to 12 times a day.

  The project is to research suitable sensor network hardware and protocols for this application, together with simulated results of the behaviour of your protocol to support your recommendations.

  Background: Thorstensen, B., Syversen, T., Bj¸rnvold, T., and Walseth, T. 2004. Electronic shepherd - a low-cost, low-bandwidth, wireless network system. In Proceedings of the 2nd international Conference on Mobile Systems, Applications, and Services (Boston, MA, USA, June 06 - 09, 2004). MobiSys '04. ACM, New York, NY, 245-255. DOI= http://doi.acm.org/10.1145/990064.99009

• Dependable Data Collection and Management in Sensor Networks

  Obtaining dependable spatial-temporal data is crucial for informed decision making in many fields from environmental science to health care. Sensor networks offer new technology for gathering such data.  However, the almost universal experience of sensor network deployments over the past decade is of unacceptably low data yields, and “data graveyards” of observation data that proves hard to use once collected. This project will research novel methods for data collection and management. The project is a joint work with Chris Huebner at University Mannheim.

Prof Amitava Datta

• Computer graphics
• Visualization
• Mobile and wireless computing
• Bioinformatics
• Parallel computing

Dr Rowan Davies

• Programming languages
• Logical foundations

Asst/Prof Tim French

• Specifying and refining security policies

  The project would consider the implementation of tools to support the formal reasoning about security policies. Particularly, we would be interested in representing the knowledge of a group of agents, and how these agents can evolve there knowledge over time. The project will involve implementing known algorithms to automatically determine the correctness of such policies.

• Auctions for autonomus agents

  This project would investigate the design of online auctions and the agents who would bid in these auctions. The aim is to provide an auction system that provides optimal results for both buyers and sellers, where the buyers are simple autonomous agents.

• Automated collection of meta-data from software repositories (with A/Prof Mark Reynolds)

  Software development companies often maintain repositories of software modules they have previously used. To efficiently reuse these modules they must have sufficient meta-data about the modules purpose, and this meta-data is often lacking. This project would look at methods to automatically generate such meta-data.

• Algorithms for model-checking (with A/Prof Mark Reynolds)

  Model-checking is the process of verifying that an implementation satisfies some formally specified property. This project would investigate the development of efficient algorithms for model-checking complex properties. This is a very difficult problem and it is not expected that a complete solution will be produced.

• Reasoning about Trust

  Trust is an integral concept in many areas such as security (do we trust someone is who they say they are), data management (do we trust a database to keep our personal details private) and sensor networks (do we trust that the data from a sensor is accurate). In many of these applications trust is dealt with in an ad hoc manner. This project would look at designing theories for reasoning about trust in a general sense. We would be interested in designing algorithms to answer questions such as "how many sensors do I need to trust in order to infer property p?", or "Can I trust agent Alice, and not trust agent Bob at the same time?". The project would establish a common semantic for trust based problems and design some simple algorithms for reasoning about trust

Assoc/Prof Bruce Gardiner

• Computational systems biology
• Cancer research
• Tissue engineering
• Clinical applications

Assoc/Prof Eun-Jung Holden

• Full Tensor Gradient (FTG) Data Visualisation for Mineral Exploration

  This project is to develop an effective visualisation technique for full tensor gradient data with a specific aim to support the delineation of anomalies associated with buried causative masses for mineral exploration.

  Further information on this project is provided in the project document.

  Assoc/Prof Eun-Jung Holden works for the Centre for Exploration Targeting (CET) in the Faculty of Natural and Agricultural Sciences.

Assoc/Prof Du Huynh

• Eye and gaze tracking

  Detection and tracking of eyelid and gaze direction has many important applications. For instance, a camera (or two cameras) can be mounted on the dashboard of a vehicle to detect drowiness and inattention of the driver; eye gaze can also be used as input for a disabled person to control the direction of movement of the wheel chair. In this project, you will investigate and implement advanced computer vision tracking techniques to detect and track the iris and eyelid in real-time.

• Tracking Multiple Humans in Video Sequences

  The tracking of multiple humans in video sequences is important in many computer vision tasks, such as video surveillance, event inference, activity classification and analysis. A successful tracking technique will require the modelling of background that can adapt to lighting changes and the modelling of human shape. This project will have a lot of fun to those students interested in capturing and processing video data. Detailed literature review on related techniques in background and human shape modelling must be included in your thesis. The tracking will be assumed to be done off-line so implementation in either C or Matlab will both be acceptable.

• Segmentation of moving objects from static background

  Motion segmentation is a means of separating moving objects from a static background in a captured video sequence. To simplify the project, we assume that we have an overhead camera mounted directly above the 3D scene. Such a set up helps to minimize occlusions. The project will involve studying existing dynamic background modelling techniques. The scope of the project is flexible.

• Outlier and Motion Detection using PCA

  Principal Component Analysis (PCA) has been widely used for dimension reduction in many problems. In the computer vision area, it has also been used for the representation of shape, motion, and appearance. PCA works on analysing the covariance matrix of the data to extract the principal components. This can be done either via the singular value decomposition (SVD) or eigen-decomposition of the covariance matrix. This process is rather expensive and so an iterative approach is often more desirable. A drawback of PCA is that it cannot handle the  presence of outliers in the data. In this project, a robust PCA method proposed by Torre and Black will be studied and implemented for the detection of outliers in data and for motion segmentation from video images. 

• Automatic Aging of Human Faces

  Automation of human face aging has many applications: prediction of facial appearance of wanted/missing persons; improving robustness of human face recognition; computer animation. In this project, the focus is on the last application. Given a face image, our aim is to age the face by N number of years. 3D synthetic face data may also be used. The project has a flexible scope. Some technical references can be found here.

  The student taking on this project is expected to be able to work with minimal guidance.

Dr Joel Kelso

• Visualisation of Influenza Epidemic Transmission

  Supervisors: Prof. George Milne, Dr Joel Kelso Our research group has constructed a program for simulating the spread of influenza through the population of a town of approximately 30,000 people. One of the outputs of the simulator is a complete trace of infection events. The goal of this project would be to present the history of the epidemic in a graphical form by overlaying information about infected individuals and their movements on a map of the simulated area.

• Construction of Virtual Communities for Influenza Epidemic Simulation

  Supervisors: Prof. George Milne, Dr Joel Kelso Our research group has constructed a program for simulating the spread of influenza through the population of a town of approximately 30,000 people. One of the inputs to the simulator is a a "virtual community", which describes a set individuals, households, workplaces and schools and the connections between them. The goal of this project will be to develop and implement a robust algorithm for constructing plausible virtual communities using census, survey and other data.

Asst/Prof Wei Liu

•  

• Title: Intelligent Talking Toy

    A couple of projects will be offered under this theme, which is part of an ARC Linkage Grant with RMIT University. Main research directions at the moment would be: 
  • Build interactive toy module plug-ins to understand ways of exposing the existing conversation manager and the knowledge base for flexible interactions,  
  • Investigate differrent ways of representing knowledge so the conversation can be extensible and fluid between topic changes.  
    A prototype conversation manager is available in Java.  

• E-bookworms: ontology learning and maintenance through text mining

  • See the E-newsworms in action by following this link: a demo of Wilson Wong's PhD work
  • Using Natural Language Processing and Statistical Analysis to automatically discover topics and concepts and the relationships between them using the WWW as a large knowledge base. Current project is to generate ontologies for chemical engineering risk management
  • Using Social Network Analysis and Temporal-Spatial Logic to understand and visualise changes of concepts across geographical regions during certain time period.

• Smart Home for aged care: a service-oriented ubiquitous computing environment

  • To build a context-aware computing environment using existing service-oriented hardware components for smart homes (using the Atlas Platform).
  • Intention recognition by tracking residents’ activities in the smart home and their interactions with robots and other smart devices.
  • Virtualising the smart-home into the Second Life virtual space.

• Cognitive agents and robots

  • Mapping low-level sensory input to high-level representation for communicable intelligence.
  • Protocols that support flexible and robust agent communication and coordination Programming and reasoning paradigms (e.g. BDI) to enable agent teamwork and service composition, problem domain (Simulation League and Standard Platform League in RoboCup).

Love of programming is essential for these projects. A varity of languages are used, Web Scripting (Perl, Python, Php), XML, C++, Java ...

Assoc/Prof Cara MacNish

• Computational Intelligence Techniques for Navigation

  The aim of this project is to investigate the use of evolutionary and allied dynamical optimisation techniques for navigating a terrain, for example an autonomous vehicle following a road or path.

• Driving a Racing Car Simulator using Visual Data

  Similar to the above, but focussing on the TORCS racing car simulator. Current attempts to learn bots for the simulator rely on sensor data available to the bots. This is very different to the data available to human players. The aim of this project is to learn to drive using the visual data available to humans.

• Investigation of Fitness Landscapes for Text Location in Images

  Location of distinctive items in images could be of great use to people with visual impairments as well as in robotics. This project looks at using a range of functions or transforms to generate landscapes from image features suitable for robust high-dimensional evolutionary style searches for object (particularly text) location and tracking.

• Computational Intelligence Techniques for Pose Estimation

  Like the above project this one is concerned with location and tracking using CI search techniques, however this project focuses on fitting multiple degree of freedom "avatars" to human pose estimation and tracking.

• Compositonal Pattern Producing Networks for Object Identification and Tracking

  Rather than using traditional functions and transformations for feature extraction, this project investigates the use of new "neural network" like technique for pattern generation and matching, called compositional pattern producing networks.

• Computational Intelligence Techniques for Searching Video Footage

  The ability to find specific items in video footage is important in a range of applications from security to web searching. The task is often made more difficult by the poor quality of footage and issues such as occlusion and deformation. The aim of this project is to investigate the use of Computational Intelligence search techniques to develop robust approximate approaches for searching difficult video.

• Investigating the AdaBoost Approach to Learning

  Whereas many traditional machine learning techniques seek to generate a single, high accuracy decision or classification from the data, the AdaBoost approach instead generates classifications from a compound sequence of many simple or less discriminating classifiers. It has been argued that this approach can result in improved performance. This project aims to test this approach in a practical domain such as those above. 

Assoc/Prof Chris McDonald

Chris has recently taught in the areas of computer networking, operating systems, computer & network security, computer architecture, distributed systems programming at The University of Western Australia and Dartmouth College and, together with these areas, his research interests include network simulation, ad-hoc & mobile networking, and Computer Science education.

• Further information is available on Chris's web page
  

W/Prof George Milne

• Modelling Complex Systems

  We utilise cellular automata and process algebraic formalisms to model complex, possibly chaotic systems, constructed out of many, simple, locally connected components. This approach has been applied to urban traffic flow modelling and simulation with success being judged by benchmarking our simulations against data sampled from the physical world. Current projects in this area include bushfire, disease spread epidemics and crowd dynamics modelling and simulation. Significantly, models are created in the Circal process algebra, a process algebra originally developed to describe and analyse the behaviour of digital logic. The applicability of Circal as a powerful descriptive medium for distinct classes of concurrent systems has been demonstrated by its application to a wide variety of application domains, including fluid flow and cardiac timing models.

• Simulation of Plant Disease Dynamics

  A plant disease simulation platform will enable various sources of plant health data and emergent pathogen or pest (EPP) biology to be utilised in a manner that will allow timely and cost-effective decisions to be made by BioSecurity managers and scientists from first detection and as the incursion evolves. At present most existing epidemic models utilise differential equations to predict outcomes and do not take into account spatial factors such as landscape effects, variable population density and explicit movement of the EPP. These models assume populations are closed and well mixed; that is, host numbers are constant and individuals are free to move wherever they wish. For the development of realistic landscape-influenced models any project must incorporate spatial information to reflect the heterogeneous environment found in the incursion zone. An alternative to using deterministic differential equations is to use a two-dimensional grid of interacting automata with each automaton modelling a sub-population at a given location. Automata interact, capturing the dynamics of an EPPs mobility. Appropriate spatio-temporal modelling techniques and simulation software will be developed to permit prediction of EPP spread over the landscape through time. This technology will build on methods developed to simulate the spread of human and animal diseases (pandemic influenza, foot and mouth disease, classical swine fever) and determine the efficacy of applying alternative eradication, containment and control strategies.

PhD Scholarships Available

Plant BioSecurity PhD and Honours Scholarships are available in the area of Computational Epidemiology and Disease Spread Modelling and Simulation within the School of Computer Science and Software Engineering at the University of Western Australia, under the guidance of Professor G J Milne.

Dr Peter Pivonka

• Computational systems biology
• Cancer research
• Tissue engineering
• Clinical applications

Prof Mark Reynolds

• Deveoping ways of managing the extraction of the purpose of software modules held in company repositories.

  With Tim French and Perth IBM software development lab.

• Formal specification of cooperative behaviour between robots.

  With Dr Wei Liu
• Our research group would be interested in an efficient method for representing and manipulating populations of organisms stratified by genotype. Some sort of multi-set is quite possibly appropriate, with the information for individual genotypes being {g, n} where g is a description of the genotype and n is the number of individuals with that genotype. It is notable that total populations may be very large (millions), that the total number of possible genotypes increases rapidly with the number of genes being considered (G³ where G is the number of genes, for diploid genotypes with 2 alleles per gene and no linkage), but that some genotypes will often be much more frequent than others. With Dr Art Diggle, Department of Agriculture, Western Australia.
  

• Temporal-spatial reasoning with incomplete information

  In this case weather radar images, to predict time of rain front hitting a certain point within the metro area. See current radar images here. There is a brief introduction to the field here. Joint supervision with Dr Du Huynh.

• Generation (via genetic programming) of simple rules for approximate solutions in complex situations

  Eg, route planning in Rogaining.

• Natural(ish) Language Querying of Simple Temporal Situations

  Given a model of several objects moving about through several locations, allow a user to use quite natural language temporal expressions (eg, will move into, did move out of, stayed until, has been there since) to query the model. See some recent related work.

• Detecting Unusual Events In Surveillance Scenes

  An automatic surveillance system requires detection of events that are unusual to the known environment. Unusual events may include a sudden change of motion speed within the scene, or detection of suspicious motion of an individual. Use a formal temporal logic language to specify what higher order events count as suspicious. Joint supervision with Dr Eunjung Holden : see more here.

• Model Checking systems against specifications with hidden (ie quantified) propositions

  Joint supervision with Tim French: see the model checking site at CMU

• Model Check

  Real-time systems modeled as networks of timed automata. See the Uppaal system.

• Automated production of test suites for data structure software.

• Reasoning about systems of Agents

  Various projects to do with logical formal reasoning about properties of systems of agents. (Jointly supervised by various combinations of Dr Wei Liu, Tim French and A/P Mark Reynolds). Contact us for details and/or discussions.

• Software Requirements Prioritization with Chunks.

  with Terry Woodings

• Visualization of Temporal Properties or Queries

  see here for some interesting related links

• Modelling local wind or tidal behaviour using cellular automata.

Prof Tim Sercomb

 Dr Ferdous Sohel

• Shape and texture-based quality metrics for bit-rate allocation for 3D video

  
  With Winthrop Professor Mohammed Bennamoun

  Summary: Current forecasts on multimedia communication predict an aggressive global growth. The transition from single medium communication (e.g., telephone conversation) to interactive multimedia with 3D audiovisual content (e.g., Internet Protocol Television, multimodal teleconferences) encompasses the capability of reaching more complete and natural forms of expression and communication. In an Australian context, this growth is expected to be accelerated due to the installation of the National Broadband Network. Quality of Service (QoS) in multimedia systems is an integral part of Quality of Experience (QoE). Perceived quality is one of the most fundamental goals when developing multimedia technologies, products and services. Therefore, an approach for reliable assessment of next generation multimedia services needs to be based on effective metrics measuring the subjective quality.  

  The title of the proposed research is: “shape and texture-based quality metrics for bit-rate allocation for 3D video”. Video objects are defined by shape, texture, motion, and depth. While shape is the most important visual cue of an object, texture takes the most amounts of resources for representation. Conversely, there is a trade-off between resource requirements (bit-rate, computational power) and quality (visual and objective). Besides, for different devices and QoS levels, requirement and availability of resources vary: e.g., a video to be displayed on mobile and HDTV screens require different objective qualities to attain the same level of perceived quality. The shape quality metrics are usually in spatial domain while the texture metrics are in feature domain. So there is a niche for the investigation of a new quality metric that is in the joint shape-texture domain and resembles the perception of human visual systems. In order to accomplish this, we will formulate the problem by introducing penalty parameters to shape and texture distortions and develop algorithms to automatically tune the parameters for an anticipated visual perception. Efficient distance estimation techniques will be investigated to obtain the distortion between 3D surfaces, as well as the effects of mesh decimation.

  Moreover, quality metrics, used in encoding and streaming, broadly accounts for the efficiency and accuracy of the system. Therefore, we will specifically focus on the investigation of new joint shape-texture based quality metrics coupled with efficient rate-distortion optimal coding schemes for current and new trends in multimedia. We will also utilise the correlation between shape and texture (e.g., shape-adaptive texture) so to ensure better encoding performance.

  Anticipated time requirement: 20 hours per week.

• Revocability of biometrics using non-invertible feature transforms

  
  With Winthrop Professor Mohammed Bennamoun

  Background: Reliable user identification and authentication are becoming increasingly important tasks in the electronic-enabled world. Biometrics (fingerprints, face, ear) based systems are becoming more popular day by day. However, a major problem with biometrics arises when the data associated with biometrics have been compromised. For instance, when an ID or password is lost, it can be changed as often as required. But when the biometrics data are lost, the relevant biometrics cannot be changed. This results in another research challenge, the so called revocability of biometrics. Derivatives of biometrics data will be used instead of the original biometrics data so that the biometric data are protected.

  Literature: BioHashing is one of the common approaches of revocable biometrics [1]. BioHash combines the biometric template (multimodal) with user-specified tokenised random number (TRN) to produce a subset of biometrics (or biometric features). However, BioHash is actually based on the sole use of TRN and that the random mixing process will destroy the optimality and efficiency of the biometrics system [1]. Moreover, according to Zhang et al. the noninvertible process may become ‘invertible’ to the extent that the matcher may be fooled [2]. Therefore, new methods for biometric revocability need to be investigated.

  Methodology: In this project, we consider the revocability of biometrics to be achieved though the introduction of an intentional, repeatable distortion to the biometrics signal based on a selected transform. There is a large domain of noninvertible transform functions in the literature such as distance transform, block scrambling and mapping, noninvertible geometric transform and noninvertible Gabor transform. These functions will be investigated in this research. The biometric signal will be distorted in the same fashion at each presentation, for enrolment and for every authentication. With our approach, every instance of enrolment will use a different transform thus rendering cross-matching impossible. Furthermore, if one variant of the transformed biometric data is compromised, then the transform function can simply be changed to create a new variant (transformed representation) for reenrolment as, essentially, a new person. In general, the distortion transforms are known to be noninvertible. So even if the transform function is known and the resulting transformed biometric data are known, the original (undistorted) biometrics cannot be recovered. In this project, we will use a combination of geometric transforms and Gabor transforms to secure noninvertibility. We further expect to investigate the effects of the distortion transforms when they are applied in either data level or feature level in order to analyse the comparative study.

  1. References:
  2. A. Teoh, Y. Kuan, and S. Lee, “Cancellable biometrics and annotations on BioHash,” Pattern Recognition, 41(6): 2034–2044, 2008.
  3. K.-H. Cheung, A. Kong, J. You, and D. Zhang, “An analysis on invertibility of cancelable biometrics based on biohashing,” in Proc. Int’l Conf. Image Science, Systems, and Technology: Computer Graphics, Las Vegas, Nevada, USA, June 2005, pp. 40–45.
  Anticipated time requirement: 20 hours per week.

• Melanoma prognosis based on 3D imaging and epidemiological risk information.

  
  With Winthrop Professor Mohammed Bennamoun

  Summary: Melanoma is the deadliest of all skin cancer types. However, if detected at an early stage, skin cancer has a very high cure rate, and requires rather simple and economical treatments. On the contrary, when detected at a late stage, cancerous lesions have very high morbidity and mortality rates, and an extremely high cost associated with accurate diagnosis for the necessary treatment. Therefore, an automatic mechanism for early detection of melanoma is crucial. There are three main techniques that are used for an automatic classification of melanoma: (i) the ABCD(E) visual diagnostic system, (ii) epiluminescent microscopy (ELM), and (ii) physics based skin modelling. The ABCDE system is one of the earliest and most effective strategies. Much research has been done on ABCDE using 2D images of the suspected skin-lesions. 3D imaging has been used in skin lesion classification in the recent years. The addition of 3D data was motivated by the experience of clinical dermatologists, who touch lesions as part of their examinations. With the addition of depth data, skin shape properties can also be extracted that are potentially of benefit in the classification process. Moreover, the depth information will improve the skin lesion segmentation accuracy. The aim of this research project is to investigate a 3D shape and colour based image analysis system for automatic melanoma prognosis which falls in the class of ABCDE. 3D data acquisition, 3D segmentation of the wound from the lesions, and 3D feature extraction from the segmentation reside at the core of this system. The Minolta VI910 scanner available at the School will be used to collect the scans of the suspected skin lesions. Various features required by the ABCDE strategy will then be determined from the data. Both feature information and matching score of ABCDE will then be fused together and the fusion results will be fed to a 3D imaging based classifier for classification of the lesions into melanoma classes in terms of their severity.

  Minimum time requirement: 20 hours per week.

Assoc/Prof Nick Spadaccini

• Medical imaging
• Computer vision

Assoc/Prof Lyndon While

• UWA Sports Staff Scheduling

  The UWA Health and Fitness Centre is a dynamic and exciting exercise environment. This is reflected in the staffing arrangements, and supporting this dynamic environment poses a significant challenge in staff scheduling. Scheduling a large number of staff with diverse skills and varying personal requirements to a range of classes, times, and venues is a non-trivial task which is difficult for humans to perform well by hand.

  This project will use evolutionary algorithms to develop an optimisation tool that will perform the following tasks.

  • Construct a staff allocation optimised for multiple criteria, such as matching staff skills to classes, staff personal requirements, and resource usage.
  • Fine-tune the initial allocation in the face of changing requirements, for example staff illness or resource unavailability.
  • Present a variety of user-friendly interfaces to stakeholders with different usage modes.
  With success the system could be extended to other parts of the UWA Sports organisation such as the Front Desk, Programs, and Recreate.
  

Prof Michael Wise

• Prof Wise's current research focus is on bioinformatics, following many years in mainstream computer science research, particularly language primitives for parallel computation and automated plagiarism detection. Prof Wise's bioinformatics research interests are diverse, ranging from keyword clustering of proteins based on text mining, through analysis of low-complexity and aggregation-prone proteins, to simulation of metabolic and other chemical reaction pathways. His philosophy is that bioinformatics is pure and simply the application of computer technology to biology, generally based on molecular-level data.

  Prof Wise is a joint appointment between the Schools of Biomoledular, Biomedical & Chemical Sciences, Computer Science & Software Engineering and in the Faculty of Natural and Agricultural Sciences.

• A Novel Method for Building Phylogenetic Trees

  Phylogeny is the study of the relatedness of species. The way this is done these days is through the computational analysis of genes in living organisms (after all, we can't go back in time to track speciation events as they happened). The phylogeny of organisms is often depicted as phylogenetic trees and there is a considerable literature on how best to create such trees. Most methods take as input data from a single gene or protein sequences. That is, the same gene is found in all the species of interest and then compared to build the tree. The problem with this approach is that it assumes that the gene is "typical" and that evolutionary pressures have acted in the same way across all the species to shape that gene. A second problem is to find a gene that is both ubiquitous and conserved in its function, but with sufficient variability to differentiate the various species possessing that gene. In this project you will create an application which extracts data generated by an existing genome analysis application as it traverses whole bacterial chromosomes, and then, after normalising the elements of the data vectors, try different methods for building phylogenetic trees from the data. In other words, rather than trying to find the idea gene around which to build a tree, this method will use all the data available in chromosomes or, by extension, entire genomes.

• Computing the Genetic Basis for the Sit-and-Wait Hypothesis of Bacterial Pathogenicity

  The Sit-and-Wait Hypothesis (Walther & Ewald 2004) relates the environmental durability of non-vector-borne microbial pathogens to their pathogenicity. Non-vector-borne pathogens, such as E. coli are passed directly from host to host, rather than being transmitted by another vector species, such as mosquitos. (For example, Ross River Fever, seen around Perth, is mosquito borne, while influenza is transmitted from person to person. Both, however, are viruses.) In the hypothesis, durability - the ability to survive the stresses associated with existing for a period outside a host - is, in effect, a cofactor for pathogenicity, in concert with the necessary presence of virulence factors. That is, without an assortment of virulence factors, a microorganism is unable to colonise a host, but if the microorganism is labile, i.e. not able to withstand desiccation, the pathogen cannot afford to be too virulent sick because an immobilised infective host is unable to move and thus unable to spread the infection, and the infection will die out.
  

International Centre for Radio Astronomy Research

• The International Centre for Radio Astronomy Research is offering three projects involved in galaxy formation simulations.These require a massive amount of computing power. Current code requires significant HPC and and use MPI for internode communications.  Simulations typically run for months at a time and require many terrabytes of storage with sophisticated checkpointing.
• 
  
  
• If you are interested, please download this document and contact Kevin Vinsen or alternatively visit http://www.icrar.org/ for further information.

• Project 1

  TBA

• Project 2

  TBA

• Project 3

  TBA