Performance metrics and methods for tracking in surveillance
Tim J. Ellis · Research Repository (Kingston University London) · 2002
Performance evaluation has become an increasingly important feature of video surveillance systems, as researchers attempt to assess the reliability and robustness of their operation. Although many algorithms and systems have been developed to address the problem of detecting and tracking moving objects in the image, few systems have been tested in anything other than fairly ideal conditions. In order to satisfy the requirements of a real video surveillance task, the algorithms will need to be assessed over a wide range of conditions. The aim of this paper is to examine some of the main requirements for effective performance analysis and to examine methods for characterising video datasets.