Browse by author
Lookup NU author(s): Kumanan Subramaniam,
Professor Satnam Dlay,
Dr Claire Rind
Full text for this publication is not currently held within this repository. Alternative links are provided below where available.
This paper presents an algorithm to solve the problem of looming motion segmentation as a part of a vehicle tracking system. The basic algorithm, uses the Gabor and Mallat Wavelet transforms to produce motion segmentation objects as an input to the present algorithm. When an object is looming in a scene, it is growing in all directions and treated as different objects by the motion segmentation algorithm. The edges of these different parts of the object are connected locally in the scene. By applying the idea of edge connectivity, based on mathematical morphology, the segmented parts of the object can be connected together. This improved algorithm can detect and produce segmented results for both looming and translational motion. Applying Boolean algebra for connectivity decision, makes the implementation of the algorithm simple and increases the speed of computation. It also copes well with multiple objects looming in the scene. Reliability of the connected objects is increased by cross correlating the Mallat information (edge detection) and final result (connected segmented objects). The computational time is reduced further since the images are binary and are processed using morphological operators.
Author(s): Rind FC; Dlay SS; Subramaniam K; Shukla S
Publication type: Book Chapter
Publication status: Published
Book Title: Advances in Physics, Electronics and Signal Processing Applications
Print publication date: 01/01/2000
Publisher: World Scientific Engineering Society
Library holdings: Search Newcastle University Library for this item