Output details

Learning scene context for multiple object tracking.

<23>The paper contributes to multiple object tracking in video of a noisy environment by giving the first demonstration of application of the Probability Hypothesis Density (PHD) filter to provide environmental modelling. The filter distinguishes likely locations for births of targets and hence helps distinguish between births and clutter, leading to a significant increase in efficiency and accuracy. The work was a key outcome of EPSRC First Grant (EP/D033772/1) and led to invited presentations at four international advanced schools: Spring School on Surveillance, 2011; Machine Vision Autumn School, 2010, Signal processing for security systems Summer School, 2010; Intermedia Summer School, 2010.