When: Thursday 24th of March, 1PM AEDT
Where: The talk will be partially presented at the Rose Street Seminar area (J04) and partially online via Zoom, RSVP here.
Speaker: Ahalya Ravendran
Title: Burst Imaging for Light-Constrained Reconstruction
State-of-the art reconstruction methods perform well under good lighting conditions but fail in low light due to loss of visual information. This seminar establishes the viability of adapting burst imaging to improve robotic vision in low light. We show that the use of burst imaging for light-constrained reconstruction outperforms existing conventional methods of doing so, yielding more accurate models and camera pose estimation. We also show burst imaging allows reconstruction to operate in conditions where it previously could not. In addition, we introduce our latest burst feature detector which jointly search scale-motion invariant features within a scale-slope space. We show in our preliminary experiments that our burst feature finder can operate on lower signal-to-noise ratio (SNR) regime than the conventional methods. This work enables robots to operate in conditions where they could not operate before.
Ahalya Ravendran is a final year PhD student in the Robotic Imaging group at SIRIS/ACFR, exploring novel ways a robot can see in a visually challenging environment. Her research interests include computational imaging, robotic vision, and feature-based reconstruction. She loves an image that a robot would generally hate.