A novel zoom invariant video object tracking algorithm (ZIVOTA)
Date
2003
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Advanced surveillance system should have the ability to track targets that have non-rigid shape and structure/size. A limitation in the field of real-time target tracking is the high computation load, so the tracking algorithms are difficult to be implemented on an ordinary personal computer without any special-purpose hardware component. This thesis presents a real-time zoom-invariant video object tracking algorithm (ZIVOTA). It can be used to track human or rigid-shaped object. ZIVOTA can provide information of object-of-interest while performing tracking, which enables automatic visual system control. It employs the same set of features to track both the position and size of a moving object through frames. Time-consuming processes like feature detection and affine basis set selection will be done only when it is necessary. Computational load is largely reduced. By using affine motion model, this technique is fundamentally zoom-invariant, it can deal with translation, rotation, and scaling of objects. Since the gaze point (representing the object location) and object boundary points (representing the object size) are virtual, ZIV OT A is able to handle partial occlusion.
Description
Bibliography: p. 82-88
Keywords
Citation
Wei, Y. (2003). A novel zoom invariant video object tracking algorithm (ZIVOTA) (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. doi:10.11575/PRISM/20605