Skip to main content


Photo of Prof Jacqueline Christmas

Prof Jacqueline Christmas

Associate Professor of Machine Learning (E&R)


Telephone: 01392 723039

Extension: (Streatham) 3039

Research interests

Machine learning for intelligent image and video understanding.   Bayesian modelling and variational approximation.   Sea wave prediction and quiescent period prediction.   Maritime applications of Bayesian modelling and simulation.

Current projects

Sea Wave Prediction
Working with Prof. Michael Belmont, Hon. Prof. Dr Bernard Ferrier and Dr Mustafa (Fass) Al-Ani to make significant improvements to the safety of maritime launch and recovery operations by providing short-term predictions of the profiles of the waves. We are pioneering research into Quiescent Period Prediction (QPP) which aims to predict when short periods of relative calm are about to occur. This has the potential to allow a range of wave critical marine operations to be safely carried out at considerably larger wave amplitudes than would otherwise be possible. Our work has been funded by the EPSRC (ref EP/N009142/1) and we are currently working directly with the Royal Navy and MOD.
PhD student Antonis Loizou has recently passed his PhD viva, subject to minor corrections, working on methods for measuring the sea surface from radar and video.

Multi-Light Imaging
Together with PhD student Matthew McGuigan and the Metropolitan Police, we are working on a multi-light imaging method for extracting images of latent fingerprints from difficult surfaces, such as lighbulbs. The new technique is described in "Remote Extraction of Latent Fingerprints (RELF)", which demonstrates how good the results are. RELF is fully automated, contactless and chemical-free, meaning that the original latent prints remain available for other forms of forensic analysis.
Matthew and I are also working on a method for enabling Reflectance Transformation Imaging (RTI) to be carried out on specular surfaces.

Real-time Bayesian inference for non-stationary systems
Noisy sensors may be in situ for a considerable (possibly effectively infinite) period of time, and the system they are sensing may be statistically non-stationary. In "Non-stationary, online variational Bayesian learning, with circular variables" I introduce a means of continuously learning from such time-series streams, potentially in real time.

Projects in the pipeline
I have further projects in the pipeline, with partners in: Babcock and BMT; QinetiQ; the Royal Navy and MOD; the Metropolitan Police; and the Austrian Institute of Technology and a number of police and forensic institutes across Europe.