DH 1st floor SR

Problem #1: (marker-less scaling) Poikos ltd. has created algorithms for matching photographs of humans to three-dimensional body scans. Due to variability in camera lenses and body sizes, the resulting three-dimensional data is normalised to have unit height and has no absolute scale. The problem is to assign an absolute scale to normalised three-dimensional data.

Prior Knowledge: A database of similar (but different) reference objects with known scales. An imperfect 1:1 mapping from the input coordinates to the coordinates of each object within the reference database. A projection matrix mapping the three-dimensional data to the two-dimensional space of the photograph (involves a non-linear and non-invertible transform; x=(M*v)_x/(M*v)_z, y=(M*v)_y/(M*v)_z).

Problem #2: (improved silhouette fitting) Poikos ltd. has created algorithms for converting RGB photographs of humans in (approximate) poses into silhouettes. Currently, a multivariate Gaussian mixture model is used as a first pass. This is imperfect, and would benefit from an improved statistical method. The problem is to determine the probability that a given three-component colour at a given two-component location should be considered as "foreground" or "background".

Prior Knowledge: A sparse set of colours which are very likely to be skin (foreground), and their locations. May include some outliers. A (larger) sparse set of colours which are very likely to be clothing (foreground), and their locations. May include several distributions in the case of multi-coloured clothing, and will probably include vast variations in luminosity. A (larger still) sparse set of colours which are very likely to be background. Will probably overlap with skin and/or clothing colours. A very approximate skeleton for the subject.

Limitations: Sample colours are chosen "safely". That is, they are chosen in areas known to be away from edges. This causes two problems; highlights and shadows are not accounted for, and colours from arms and legs are under-represented in the model. All colours may be "saturated"; that is, information is lost about colours which are "brighter than white". All colours are subject to noise; each colour can be considered as a true colour plus a random variable from a gaussian distribution. The weight of this gaussian model is constant across all luminosities, that is, darker colours contain more relative noise than brighter colours.

In this seminar we discuss the gas dynamics of chimneys, solar updraft towers and energy towers. The main issue is to discuss simple fluid dynamic models which still describe the main features of the mentioned applications. We focus first on one dimensional compressible models. Then we apply a small Mach number asymptotics to reduce to complexity and to avoid the known problems

of fully compressible models in the small Mach number regime. In case of the energy tower in addition we have to model the evaporation process.

Finally we obtain a much simpler fluid dynamic model which allows robust and very fast numerical simulations. We discuss the qualitative behaviour and the good agreement with expermental data (in cases such data are available).

'Pattern-of-life' is a current buzz-word in sensor systems. One aspect to this is the automatic estimation of traffic flow patterns, perhaps where existing road maps are not available. For example, a sensor might measure the position of a number of vehicles in 2D, with a finite time interval between each observation of the scene. It is desired to estimate the time-average spatial density, current density, sources and sinks etc. Are there practical methods to do this without tracking individual vehicles, given that there may also be false 'clutter' detections, the density of vehicles may be high, and each vehicle may not be detected in every timestep? And what if the traffic flow has periodicity, e.g. variations on the timescale of a day?

The computational analysis of a mathematical model describing a complex system is often based on the following roadmap: first, an experiment is conceived, in which the measured data are (either directly or indirectly) related to the input data of the model equations; second, such equations are computationally solved to provide iconographic reconstructions of the unknown physical or physiological parameters of the system; third, the reconstructed images are utilized to validate the model or to inspire appropriate improvements. This talk will adopt such framework to investigate three applied problems, respectively in solar physics, neuroscience and physiology. The solar physics problem is concerned with the exploitation of hard X-ray data for the comprehension of energy transport mechanisms in solar flares. The neuroscientific problem is the one to model visual recognition in humans with the help of a magnetocencephalography experiment. Finally, the physiological problem investigates the kinetics of the kidney-bladder system by means of nuclear data.

Complex networks have been used to model almost any

real-world complex systems. An especially important

issue regards how to related their structure and dynamics,

which contributes not only for the better understanding of

such systems, but also to the prediction of important

dynamical properties from specific topological features.

In this talk I revise related research developed recently

in my group. Particularly attention is given to the concept

of accessibility, a new measurement integrating topology

and dynamics, and the relationship between frequency of

visits and node degree in directed modular complex

networks. Analytical results are provided that allow accurate

prediction of correlations between structure and dynamics

in systems underlain by directed diffusion. The methodology

is illustrated with respect to the macaque cortical network.

Feature Selection is a ubiquitous problem in across data mining,

bioinformatics, and pattern recognition, known variously as variable

selection, dimensionality reduction, and others. Methods based on

information theory have tremendously popular over the past decade, with

dozens of 'novel' algorithms, and hundreds of applications published in

domains across the spectrum of science/engineering. In this work, we

asked the question 'what are the implicit underlying statistical

assumptions of feature selection methods based on mutual information?'

The main result I will present is a unifying probabilistic framework for

information theoretic feature selection, bringing almost two decades of

research on heuristic methods under a single theoretical interpretation.

Hollow vortices are vortices whose interior is at rest. They posses vortex sheets on their boundaries and can be viewed as a desingularization of point vortices. We give a brief history of point vortices. We then obtain exact solutions for hollow vortices in linear and nonlinear strain and examine the properties of streets of hollow vortices. The former can be viewed as a canonical example of a hollow vortex in an arbitrary flow, and its stability properties depend. In the latter case, we reexamine the hollow vortex street of Baker, Saffman and Sheffield and examine its stability to arbitrary disturbances, and then investigate the double hollow vortex street. Implications and extensions of this work are discussed.

Nematic liquid crystals (NLCs) are materials that flow like liquids, but have some crystalline features. Their molecules are typically long and thin, and tend to align locally, which imparts some elastic character to the NLC. Moreover at interfaces between the NLC and some other material (such as a rigid silicon substrate, or air) the molecules tend to have a preferred direction (so-called "surface anchoring"). This preferred behaviour at interfaces, coupled with the internal "elasticity", can give rise to complex instabilities in spreading free surface films. This talk will discuss modelling approaches to describe such flows. The models presented are capable of capturing many of the key features observed experimentally, including arrested spreading (with or without instability). Both 2D and 3D spreading scenarios will be considered, and simple ways to model nontrivial surface anchoring patterns, and "defects" within the flows will also be discussed.

 When modelling the motion of a sperm cell in the female reproductive tract, the Reynolds number is found to be very small, thus allowing the nonlinear Navier-Stokes equations to simplify to the linear Stokes equations stating that pressure, viscous and body forces balance each other at any instant in time. A wide range of analytical techniques can be applied to investigate the Stokes flow past a moving body. In this talk, we introduce various Stokes flow singularities and illustrate how they can provide a handy starting point (ansatz) when trying to determine the form of the flow field around certain bodies, from simple translating spheres to beating sperm tails.

A SMEC device is an array of aerofoil-shaped parallel hollow vanes forming linear venturis, perforated at the narrowest point where the vanes most nearly touch. When placed across a river or tidal flow, the water accelerates through the venturis between each pair of adjacent vanes and its pressure drops in accordance with Bernoulli’s Theorem. The low pressure zone draws a secondary flow out through the perforations in the adjacent hollow vanes which are all connected to a manifold at one end. The secondary flow enters the manifold through an axial flow turbine.

SMEC creates a small upstream head uplift of, say 1.5m – 2.5m, thereby converting some of the primary flow’s kinetic energy into potential energy. This head difference across the device drives around 80% of the flow between the vanes which can be seen to act as a no-moving-parts venturi pump, lowering the head on the back face of the turbine through which the other 20% of the flow is drawn. The head drop across this turbine, however, is amplified from, say, 2m up to, say, 8m. So SMEC is analogous to a step-up transformer, converting a high-volume low-pressure flow to a higher-pressure, lower-volume flow. It has all the same functional advantages of a step-up transformer and the inevitable transformer losses as well.

The key benefit is that a conventional turbine (or Archimedes Screw) designed to work efficiently at a 1.5m – 2.5m driving head has to be of very large diameter with a large step-up gearbox. In many real-World locations, this makes it too expensive or simply impractical, in shallow water for example.

The work we did in 2009-10 for DECC on a SMEC across the Severn Estuary concluded that compared to a conventional barrage, SMEC would output around 80% of the power at less than half the capital cost. Crucially, however, this greatly superior performance is achieved with minimal environmental impact as the tidal signal is preserved in the upstream lagoon, avoiding the severe damage to the feeding grounds of migratory birdlife that is an unwelcome characteristic of a conventional barrage.

To help successfully commercialise the technology, however, we will eventually want to build a reliable (CFD?) computer model of SMEC which even if partly parametric, would benefit hugely from an improved understanding of the small-scale turbulence and momentum transfer mechanisms in the mixing section.