A Morphological Approach to Fish Discrimination.

Abstract

In this paper, we propose a new method of the feature extraction for fish discrimination using mathematical morphology. The conventional research of fish discrimination used parameters which relied upon the physical characteristics of shape, contour and colour. But these methods were insufficient for pract,ical application system. Other physical pararneters include the colouring patterns, speckles and the scale forms of the fish. We focused on the speckle patterns and scale forms of fish as significant ident,ifica,tion features. We tried to extract the features of Japanese Horse Mackerel (ma-aji), Pilchard Sardine (ma-iwashi) and Common Mackerel (ma-saba) by iilnpleinenting newly designed morphological filters and algorithms. The discriminatio~i system was in~plemented on the distributed pa.rallel ima.ge processing system. Experimental results show the discrimination performance is superior to the conventional methods.