New Zernike velocity moments have been developed to describe an object and its motion throughout an image sequence. These are an extended form of the orthogonal Zernike moment set and include velocity information introduced via centralised moments. Initial analysis shows that they perform well when applied to analysing gait sequences resulting in a good recognition rate and a compact description. They have exhibited promising attributes when applied to occluded data, which is reflected in the method of describing a complete temporal sequence and not single images. Further, they appear to provide measures intimately related to the moving and/or morphing shape within the sequence. Their invariance properties suggest that they will be useful in real situations where poor quality or camera zoom problems are apparent. Future work will aim to investigate performance attributes on a larger dataset, as well as their capability for generic application.
Home Contents Author index Keyword index
This document produced for BMVC 2001