Research Report - Sensor and Signal Processing
A wide variety of sensors are used in applications ranging from mining to defence and from medicine to navigation. The important information collected by such sensors often has very high data rates and may be overwhelmed by extraneous signals. With the advent of very precise analog-to-digital converters, high power computing technology and powerful digital signal processing hardware, a significant improvement in sensor performance is possible.
Many of the staff and students working in this field form a node of the Cooperative Research Centre for Sensor Signal and Information Processing (CSSIP). This is a collaborative venture between The University of Adelaide, Flinders University, The University of South Australia, The University of Melbourne, The University of Queensland, DSTO, Compaq, Telstra, RLM, and CEA Technologies Pty Ltd. The CSSIP Queensland node is active in the main research areas of radar signal processing, mining sensors and cytometrics.
A major area of research for this group is in sensors for the mining industry where high technology sensors such as radar, sonar, laser and optics are becoming increasingly important. The group has been developing a second generation (stepped-frequency) ground penetrating radar (GPR) technique for over seven years and its technology has now demonstrated significant improvements over the first generation (impulse) GPR systems.
Another current project is to use a ground-based radar sensor with interferometric processing to monitor the stability of pit walls during highwall mining. This has demonstrated an ability to measure movements to less than a millimeter and produce a map of time varying motion. The Ground Penetrating Radar and the Slope Stability Radar will be available commercially through a new company GroundProbe Pty Ltd.
A further important application of GPR is in the detection of unexploded ordnance (UXO) and plastic landmines. In this application, GPR antennas are normally placed in close proximity to the ground. The antenna and its environment interact in a complicated fashion and techniques of computational electromagnetics are needed to determine behaviour in both the near and far field of the radiator. Some targets, such as plastic landmines, are difficult to detect with a GPR and so special characteristics, such as resonant frequencies, are sought to aid detection. Software developed allows accurate and complete computer design of the radiators used in this GPR application and also provides insight into many of the phenomena observed in empirical and experimental testing of the radar with different conducting targets.
A related area of study is in synthetic aperture radar (SAR) systems which provide high resolution 2- dimensional images of the ground, irrespective of atmospheric or illumination conditions. SAR images, which may be acquired from spaceborne or ground-based systems, contain both reflectivity and phase information. Research within the group is aimed at: generation of high quality 3-dimensional images of terrain through (i) multi-baseline interferometric processing of a series of spaceborne SAR images acquired over the same terrain (ii) development of radar waveform diversity and array null steering techniques to increase the swath width of spaceborne SAR systems (iii) development of SAR image texture modelling techniques to enable segmentation/classification of the different terrain types, and (iv) development of clutter modelling and suppression techniques for ground-based ultra-wide band SAR so as to improve the detectibility of, for example, land-mines.
One major topic of research in the group that does not involve radar is cytometrics, which is the application of computer vision, image analysis, and pattern recognition techniques to the detection of cancer and neoplasia (early pre-cancer) on pathology slides. This work is developing software for the automated screening of Pap smear slides for the detection of cervical cancer and should lead to improved efficiency, accuracy and sensitivity of cancer detection in pathology laboratories. Computers can detect and measure changes too small to be visible to the human eye and can therefore give new information on the degree of abnormality. Such systems may be useful in a wide range of diseases and diagnostic tests using slides. The research currently focuses on the detection of early changes in the visual appearance of cell nuclei caused by pre-cancerous changes at a lesion nearby in the body. Work so far has produced a suite of very powerful abnormality detection algorithms using image texture in the cell nucleus as a diagnostic marker.
Associated Staff
Dr Vaughan Clarkson
Prof Stuart Crozier
Prof Tom Downs
Dr John Homer
A/Prof Peter Kootsookos
Prof Dennis Longstaff
A/Prof Brian Lovell
<%@ include file="/include/footer.inc.jsp" %>