Fish recognition method using vector quantization histogram for investigation of fishery resources

This paper proposed a novel approach to the kichiji rockfish recognition method for investigation of the fish biomass. This method does not depend on the size, position and attitude of the fish, because SIFT feature with rotation invariant is used and is represented by vector quantization histogram. The experiment results showed that our method detected 94 % of object which is the fish from 37 photographs and recognized 57 % of the fish from detected image. And our method was able to recognize kichiji rockfish from SIFT features of the fish head. However, if more than one object is piled up, our method is a possibility of mistaken recognition because features of object other than target are used for recognition.