Navigating beyond RGB: Spectral perception for Mobile Robots
Pozývame vás na prednášku zahraničného profesora Simon Hadfield na tému Navigating beyond RGB: Spectral perception for Mobile Robots. Prednáška je otvorená aj pre verejnosť.
Kedy: 12. máj 2026, 10:00 (čas sa ešte môže zmeniť)
Kde: upresníme
O čom bude prednáška:
obotic perception has historically been shaped by the limitations of human vision, with most systems relying almost exclusively on RGB imagery. Yet light is continuous, and the rich spectral signatures produced by real-world materials contain information that is largely discarded by conventional sensors and algorithms. In this talk, I will argue that moving beyond RGB, towards multispectral and hyperspectral perception, offers a powerful route to more robust localisation, mapping, and scene understanding in mobile robots.
I will present a series of recent works from my group that explore how spectral information can be integrated into practical robotic systems. This includes using sparse spectrometer measurements for highly distinctive robot localisation, learning-based spectral reconstruction methods that recover dense hyperspectral representations from limited sensing, and extending modern 3D scene representations (such as Gaussian splatting) into the hyperspectral domain to create hyper-photo-realistic models. I will also give a preview of ongoing work on online hyperspectral SLAM, where localisation, mapping, and material understanding are performed jointly during exploration.
Throughout the talk, I will highlight both the opportunities and the challenges of spectral perception, including sensing trade-offs, computational constraints, and learning in high-dimensional spaces. I will conclude by discussing how hyperspectral sensing can enable robots to understand not just where things are, but what they are, opening up new possibilities for inspection, maintenance, and embodied interaction.
O hosťovi:
Dr. Simon Hadfield, Associate Professor (Reader) in Robot Vision and Autonomous SystemsCentre for Vision Speech and Signal Processing (CVSSP) & Fellow of the Institute for People Centred AIUniversity of Surrey, Guildford, Surrey, England
Prednášku organizuje Ústav robotiky a kybernetiky FEI STU





