PhD Thesis Defence: Robotic Burst Imaging for Light-Constrained 3D Reconstruction, 24th May, 4pm

When: Wednesday 24th of May, 4pm AEST

Where: This seminar will be partially presented at the Rose Street Seminar area (J04) and partially online via Zoom. RSVP

Candidate: Ahalya Ravendran

Title: Robotic Burst Imaging for Light-Constrained 3D Reconstruction


In this talk, we introduce robotic burst imaging, a novel input scheme designed to enhance vision-based 3D reconstruction for robots operating in low-light conditions. The performance of state-of-the-art robotic vision algorithms is hindered in such environments due to the low signal-to-noise ratio. To avoid the limitation in carrying an additional light source for weight- and power-restricted robotic platforms, developing low light image processing techniques is essential. Burst imaging demonstrates one such successful low light enhancement for mobile photography. In our work, we adapt burst imaging for robotics to perform downstream tasks by leveraging platform dynamics while operating in lower light than robots could previously operate in.

We aim to improve the feature extraction stage of 3D reconstruction methods using robotic burst imaging. To this end, we introduce a scheme allowing robots to employ burst-merged images in 3D reconstruction. We then improve on this with a direct 2D + time burst feature finder and an end-to-end learning-based feature extractor. We conduct extensive testing using burst imagery captured on both a robotic arm and drones to validate our novel input scheme. We demonstrate significant improvements in low light reconstruction, culminating in an 80% improvement in convergence rate for scenes captured in millilux conditions compared with conventional imaging schemes. We also demonstrate improved localisation and model completeness.

These advancements hold great potential for various applications, such as enabling autonomous driving and drone delivery at night, facilitating efficient mining operations, and conducting behavioural studies on nocturnal animals. Our research findings pave the way for exciting opportunities in the field of vision-based 3D reconstruction, enabling robots to operate effectively in challenging low light, without a need for specialised camera hardware.


Ahalya Ravendran is a PhD student in the Robotic Imaging group at the Australian Centre for Robotics, with Dr. Donald Dansereau as her primary supervisor and Dr. Mitch Bryson as her co-supervisor. She received her BSc honours degree in mechatronics engineering from Sri Lanka Institute of Information Technology and her MSc research degree in robotics from Thammasat University, Thailand. Her research interest lies in the field of robotic vision, with a focus on imaging challenging scenes that are otherwise difficult for robots to understand.


Australian Centre for Robotics