FEATURE-EXTRACTION FOR A MACHINE-VISION-BASED SHRIMP DEHEADER

Various shrimp features were investigated for determination of the opt imum location for the removal of heads. Both morphological and spectra l features were evaluated. Morphological features included carapace wi dth to carapace length ratio, carapace length to total length ratio, a nd the skeleton ratio. The spectral feature was found effective in det ermining cutting location. Standard deviation of predictions made with the spectral feature ranged from 2.8 to 4.6 mm at a rate of two shrim p per second.