Vision Projects

ForSight: Technology to Restore Vision in Humans

(Contact: Dr. S.F. Al-Sarawi, alsarawi@eleceng.adelaide.edu.au, Dr. B.W. Ng, bwng@eleceng.adelaide.edu.au, Dr. T. Rainsford, tamath@eleceng.adelaide.edu.au)

More than two-thirds of today’s blindness could be prevented or treated by applying existing knowledge and technology. In 1997 the World Health Organization estimated that there were close to 150 million individuals with significant visual disability. In Australia, each year 10,000 people go blind, adding to the half million who are already visually impaired in both eyes. Artificial Human Vision systems utilize the perception of phosphenes as a substitute for normal vision. In many cases though, the damage is such that the blind person is no longer able to produce phosphenes and direct stimulation of the cortex is the only possible solution. Based on this and given that many people would prefer a non-invasive solution we propose a bio-inspired system that is based on how bats perform vision. Here the visual information of a person's surroundings is sensed using ultrasound waves and delivered to the auditory system via a set of headphones. Previous research into using sound to relay visual information uses camera based systems that convert image frames into sound but the bandwidth requirements severely limit its use, and difficulties with extracting features from images can degrade system effectiveness. Our proposed system is similar to a radar system in that the ultrasound source wavelength determines the resolution of objects that can be detected by this system. The higher the sound frequency, the better the resolution. Conceptually, the ultrasound source and microphones play the same role as RF emitters and receivers in a radar system, thereby enabling us to bring our expertise in radar to this problem. Initially we use of two ultrasound microphones to allow for stereo detection of the sound. In the controller unit, a smart algorithm is used to allow the conversion of these stereo signals into a mixed signal combination that is then delivered as a stereo sound to the human audio system through earpieces as shown.

Support: The University of Adelaide

Artificial Insect Vision Chips

(Contact:  Dr. D. Abbott, dabbott@eleceng.adelaide.edu.au)

The objective is to map insect vision algorithms onto VLSI to make smart collision avoidance chips. Our goal is to apply the latest neurophysiological models using contrast adaptation.

Support: Australian Research Council, Sir Ross & Sir Keith Smith Foundation, AFSOR (USA)

Millimetre-wave Insect Vision

(Contact:  Dr. D. Abbott, dabbott@eleceng.adelaide.edu.au)

This project is researching a novel motion detector utilising a millimetre-wave array front-end with signal processing that mimics insect vision. The use of passive millimetre-wave detection enables a significant improvement over optical or infrared wavelengths when rain, steam or other aerosols obscure a colliding object. This, for instance, used as a blind-spot detector, will enhance driver safety in poor weather conditions. As insect vision techniques do not attempt to process an image, but rely on tracking moving edges, the processing tasks are less hardware intensive resulting in a compact low-cost solution. We are also investigating the use of stochastic resonance to improve the detected signal to noise ratio.

Support: Australian Research Council, Sir Ross & Sir Keith Smith Foundation

Robust Motion Detection Estimation Algorithms Targeted for VLSI Technology

(Contact: Dr. T. Rainsford, tamath@eleceng.adelaide.edu.au)

A small low-cost motion detector would have widespread applications in visual control systems such as miniature unmanned aerial vehicles and collision avoidance systems. Artificial real-time vision and simple seeing systems are massively challenging because the environment greatly impacts on their performance. VLSI is ideally suited to the parallel processing seen in nature because it allows for high device integration density and implementation of complex functions. However, VLSI imposes bounds on the types of algorithms that can be implemented. This project seeks to develop and implement improved algorithms with robust outputs that are practical in terms of real time implementation in mixed analogue-digital VLSI.

Support: The University of Adelaide