Catalogue of Artificial Intelligence Techniques
Aliases: Conditional Density Propagation Algorithm
Keywords: algorithm, angles, condensation, conditional, contour, density, joint, object, probabilistic, propagation, random, real, time, tracking
Categories: Pattern Recognition and Image Processing
Author(s): Jay Kotak
The Condensation Algorithm, or to give its full name, the conditional density propagation algorithm, is a probabilistic algorithm that is used to track an objects outline in image streams in near real time.
The algorithm works by randomly sampling pixels of the current image and detecting the small changes in the distribution of pixels at successive time steps. It then uses multiple accurate hypothesis on models of shape and motion to update the displayed outline.
A brief training period will help provide optimal tracking success for the algorithm. In an uncluttered environment the algorithm alone is sufficient for this. However for a cluttered one 'bootstrapping' is recommended, typically through the use of Kalman Filtering.
It is also noted that although primarily developed for tracking curves, the algorithm could in theory be used to track joint angles instead.
For assorted video's of the algorithm in use, visit : http://www.robots.ox.ac.uk/~misard/condensation.html
- Isard, M. and Blake, A. , CONDENSTATION – Conditional Density Propagation for Visual Tracking. International Journal Of Computer Vision: Volume 29 (1998), Springer, Netherlands, pp 5-28, Available Online At: http://www.springerlink.com/content/xl887466h454318k/ And http://www.robots.ox.ac.uk/~ab/abstracts/ijcv98.html .