Timothy Farnworth


bayesian state estimation using optical flow

My research focuses on giving robot’s the ability to perceive the world the same way we do. The project’s apogee is a Bayesian estimation approach to utilising optical flow to estimate trajectory and position. The Bayesian approach provides a mathematical frame work for holistically fusing data from various sensors such as GPS, LIDAR, accelerometers, gyros, magnetometers and velocity sensors. To apply Bayesian methods, optical flow uncertainty must be characterised so that it’s information can be utilised.

optical flow measurement compression

Dense optical flow algorithms supply a plethora of information on the trajectory of a camera and the structure of the world. This information is highly correlated and has high dimensionality. I am researching methods for compressing this data to a manageable set of inducing flow vectors which describe the entire flow field by making suitable approximations.

Camera calibration

Wide field of view cameras have the ability to distinguish between a camera rotation and translation. However, planar projection models fail to capture the lens distortion of such cameras. I am currently researching the use of non-parametric spherical camera models and methods for calibration.


Our background and vision


The research team


A list of research publications


Contact us

Close Menu