nextuppreviouscontents
Next:Distributed Robot Control Software Up:Active Vision Previous:Optimal Spiral Search

PCA Background Invariance

This project deals with appearance based recognition using Principal Components Analysis with the added ability to account for varying backgrounds and non-rigid deformation. This is done using an attention operator to focus on the object to be recognised and performing PCA only on the sub-windows within the object. This allows an object to be recognized as a collection of familiar fragments that ``vote'' for it's identity. Each voting patch has a strong association with the particular object being recognized, and hence only a small number of these need be determined. This work is also being examined in the context of place recognition, for navigation.

D. Jugessur, G. Dudek
 


Annual Report

Mon Jun 26 21:22:20 GMT 2000