Flame imaging of gas-turbine relight

High-altitude relight inside a lean-direct-injection gas-turbine combustor is investigated experimentally by highspeed imaging. Realistic operating conditions are simulated in a ground-based test facility, with two conditions being studied: one inside and one outside the combustor ignition loop. The motion of hot gases during the early stages of relight is recorded using a high-speed camera. An algorithm is developed to track the flame movement and breakup, revealing important characteristics of the flame development process, including stabilization timescales, spatial trajectories, and typical velocities of hot gas motion. Although the observed patterns of ignition failure are in broad agreement with results from laboratory-scale studies, other aspects of relight behavior are not reproduced in laboratory experiments employing simplified flow geometries and operating conditions. For example, when the spark discharge occurs, the air velocity below the igniter in a real combustor is much less strongly correlated to ignition outcome than laboratory studies would suggest. Nevertheless, later flame development and stabilization are largely controlled by the cold flowfield, implying that the location of the igniter may, in the first instance, be selected based on the combustor cold flow. Copyright © 2010.

Spatio-temporal clustering of probabilistic region trajectories

We propose a novel model for the spatio-temporal clustering of trajectories based on motion, which applies to challenging street-view video sequences of pedestrians captured by a mobile camera. A key contribution of our work is the introduction of novel probabilistic region trajectories, motivated by the non-repeatability of segmentation of frames in a video sequence. Hierarchical image segments are obtained by using a state-of-the-art hierarchical segmentation algorithm, and connected from adjacent frames in a directed acyclic graph. The region trajectories and measures of confidence are extracted from this graph using a dynamic programming-based optimisation. Our second main contribution is a Bayesian framework with a twofold goal: to learn the optimal, in a maximum likelihood sense, Random Forests classifier of motion patterns based on video features, and construct a unique graph from region trajectories of different frames, lengths and hierarchical levels. Finally, we demonstrate the use of Isomap for effective spatio-temporal clustering of the region trajectories of pedestrians. We support our claims with experimental results on new and existing challenging video sequences. © 2011 IEEE.