BibTeX entry

@PHDTHESIS{200303Nils_T_Siebel,
  AUTHOR={Nils T Siebel},
  TITLE={Design and Implementation of People Tracking Algorithms
    for Visual Surveillance Applications},
  SCHOOL={Reading University},
  MONTH=Mar,
  YEAR=2003,
  URL={http://www.bmva.org/theses/2003/2003-siebel.pdf},
}

Abstract

This thesis describes the design and implementation of a people tracking module for an integrated visual surveillance system developed as part of the Framework V project ADVISOR, funded by the European Union. Starting from an earlier people tracker developed at the Universities of Leeds and Reading over the past 10 years, appropriate methods have been included and adapted to meet the special needs which exist in visual surveillance systems. The result is the Reading People Tracker. The emphasis in this work lies on the completion of the whole task: from image processing and tracking algorithms through to system design and integration. The Reading People Tracker uses a combination of four co-operating detection and tracking modules to track people in camera images. Each of the modules is of medium to low complexity to ensure realtime performance. The modules are a Motion Detector, a Region Tracker, a Head Detector and an Active Shape Tracker. Combining different modules and letting them exchange their results overcomes the limitations which individual modules have and results in a greater reliability. All four modules work directly with the image in order to reduce the complexity of the tracker. Complicated 3D models are avoided. Experiments show that the Reading People Tracker tracks individuals robustly in the presence of occlusion and low image quality. An important aspect of the Reading People Tracker is its maintainability. The software was re-engineered, making it possible to adapt, extend and improve the software with ease. This re-engineering process and its influence on the maintainability of the people tracker is analysed as a case study. A second aspect examined in the case study is the software processes which have been used to maintain the software over its lifetime of 10 years. It is shown that the diversity of software processes has a significant impact on the quality of the software. Process diversity is divided into two broad classes, latitudinal and longitudinal process diversity. The influence of both classes of process diversity on software quality is examined.