My research programme focuses on perception-aware machine vision and embedded intelligence for autonomous systems. It develops from understanding human visual perception, through efficient visual processing and embedded machine vision, towards information prioritisation and autonomous decision-making for UAVs and robotic systems.
Human Perception → Perception-Aware Machine Vision → Embedded Intelligence → Autonomous Decision-Making
Understanding which spatial, temporal, and peripheral visual information is perceptually important.
This work measured large-scale thresholds of visibility for spatio-temporal sinusoidal variations using different spatial sizes and temporal modulation rates. The resulting model describes human contrast sensitivity under modern video presentation conditions.
Based on research into the limits of human vision, this project offers a method, software, and test equipment for measuring visual system characteristics using controlled spatial, temporal, and colour stimuli.
This work contributes to streaming quality assessment by creating a large-scale dataset with video compression and transmission artefacts. The final dataset includes millions of perceptual thresholds related to video quality.
This work analyses human perception of temporal frequencies in central and peripheral vision. It presents a new method for quality measurement based on contrast sensitivity models and visibility thresholds for spatio-temporal variations on modern displays.
Using perceptual models to reduce redundant visual information before transmission, embedded processing, and AI inference.
Limited radio transmission bandwidth seriously restricts the use of wide-angle video in real-time technologies. This work combines human visual perception, embedded processing, and wide-angle vision to reduce bandwidth and enable real-time UAV and robotic operation beyond the line of sight.
Rather than only making AI models smaller, this research reduces the amount of visual information a model needs to process. It reduces training video dataset redundancy by leveraging human visual perception, improving internal computation and predictive accuracy on resource-constrained platforms.
This project developed a non-reference video quality metric incorporating psychophysical features of human visual perception, supporting real-time perceptual optimisation and stable prediction of subjective quality ratings across independent datasets.
This project develops a motion- and perception-adaptive H.265/HEVC framework for real-time UAV piloting. It combines motion-vector analysis, adaptive coding orientation, and peripheral-vision-based regional quantisation to reduce bitrate while maintaining comparable objective quality under bandwidth constraints.
Extending perception-aware processing towards information prioritisation, onboard intelligence, and autonomous decision-making for UAVs and robotic systems.
This project develops autonomous UAV systems combining onboard perception, embedded AI, visual sensing, and environmental sensing for cyclone-resilient agriculture and environmental monitoring. The research explores how autonomous platforms can prioritise relevant information, process it onboard, and support rapid decision-making under real-world computational and communication constraints.
Industry-connected research integrating UAVs, sensing, embedded systems, machine vision, and AI for infrastructure inspection and monitoring.
Research integrating sensing, embedded computing, remote monitoring, machine vision, and AI for agricultural and horticultural environments, supporting New Zealand primary-sector applications.
My next research stage focuses on visual attention and information prioritisation for autonomous systems. The goal is to allow robots and UAVs to identify decision-relevant information before full processing and respond more efficiently to important events.
Additional projects that connect this research programme to teaching and industry.
Many students view AI as an all-knowing technology and feel anxious about its impact on learning and professional futures. This project explores how lecturers can help students understand AI tools, academic integrity, and responsible adoption across disciplines.