Factors Affecting The Training Of A WISARD Classifier For Monitoring Fish Underwater
Download:
pdf
by S Hockaday, R D Tillett, L G Ross
http://www.bmva.ac.uk/bmvc/1999/papers/34.pdf
Add To MetaCart
Abstract:
A non-invasive system for monitoring fish underwater is described. A 3D point distribution model (PDM) model can be fitted on stereo images for fish examination. Currently the model fitting algorithm requires manual initialisation. Therefore experiments were carried out to investigate the usefulness of an n-tuple classifier as a tool to initiate the model fitting method automatically. The experiments were designed to identify factors that will affect the performance and the usefulness of the classifier under the requirements of the fish inspection application. Experimental results show that the classifier is a useful tool for interpreting underwater fish images and could be used as a tool to aid the application in question.
Citations
| 13 | Practical face recognition and verification with WISARD – Stonham |
| 10 | The Theoretical and Experimental Status of the n-tuple Classifier – Rohwer, Morciniec - 1995 |
| 5 | Fitting 3D point distribution models of fish to stereo images – McFarlane, Tillett - 1997 |
| 3 | A trainable n-tuple pattern classifier and its application for monitoring fish underwater – Chan - 1999 |
| 2 | Image processing for underwater measurement of salmon biomass – Chan - 1998 |
| 2 | Using stereo image pairs to measure mass in strains of Atlantic salmon, Salmo salar L – Hockaday - 1997 |
| 1 | et al. Predicting biomass of Alantic salmon form morphometric lateral measurements – Beddow - 1996 |
| 1 | et al. FishSizing and Monitoring Using a Stereo Image Analysis System Applied to Fish Farm – Ruff - 1995 |
