Logo des Repositoriums

Auflistung nach Autor:in "Maskell, Simon"

1 - 2 von 2
  • Konferenzbeitrag
    Background Modeling Using Adaptive Cluster Density Estimation for Automatic Human Detection
    (Informatik 2007 – Informatik trifft Logistik – Band 2, 2007) Bhaskar, Harish; Mihaylova, Lyudmila; Maskell, Simon
    Detection is an inherent part of every advanced automatic tracking system. In this work we focus on automatic detection of humans by enhanced background subtraction. Background subtraction (BS) refers to the process of segmenting moving regions from video sensor data and is usually performed at pixel level. In its standard form this technique involves building a model of the background and extracting regions of the foreground. In this paper, we propose a cluster-based BS technique using a mixture of Gaussians. An adaptive mechanism is developed that allows automated learning of the model parameters. The efficiency of the designed technique is demonstrated in comparison with a pixel-based BS [ZdH06].
  • Konferenzbeitrag
    Using a probabilistic hypothesis density filter to confirm tracks in a multi-target environment
    (INFORMATIK 2011 – Informatik schafft Communities, 2011) Horridge, Paul; Maskell, Simon
    In this paper, we aim to perform scalable multi-target particle filter tracking. Previously, the authors presented an approach to track initiation and deletion which maintains an existence probability on each track, including a “search track” which represents the existence probability and state distribution of an unconfirmed track. This approach was seen to perform well even in cases of low detection probability and high clutter levels, but modelling all unconfirmed tracks by a single-target search track can be problematic if more than one target appears in a sensor's field of view at the same time. To address this, we replace the search track with a Probabilistic Hypothesis Density (PHD) filter which can maintain a density over several unconfirmed tracks. A method is proposed to derive probabilities of measurements originating from targets, allowing us to confirm tracks when these probabilities reach a threshold. We observe that in so doing, we implicitly solve the track-labelling challenge that otherwise exists with PHD filters. This is shown to maintain good tracking performance for highclutter, low-detection scenarios while addressing the shortcomings of the single-target search track approach. We also show results from a scenario with obscured regions where the target cannot be detected, and show that targets can be tracked through the obscurations.