In recent years, automatic visual coral reef monitoring has been proposed to solve the demerits of manual monitoring techniques.\r\nThis paper proposes a novel method to reduce the computational cost of the standard Active Appearance Model (AAM) for\r\nautomatic fish species identification by using an original multiclass AAM. The main novelty is the normalization of speciesspecific\r\nAAMs using techniques tailored to meet with fish species identification. Shape models associated to species-specific AAMs\r\nare automatically normalized by means of linear interpolations and manual correspondences between shapes of different species. It\r\nleads to a Unified Active AppearanceModel built from species that present characteristic texture patterns. Experiments are carried\r\nout on images of fish of four different families. The technique provides correct classification rates up to 92% on 5 species and\r\n84.5% on 12 species and is more than 4 times faster than the standard AAM on 12 species.
Loading....