Discrimination of fish species with neural networks: A study on setting of color data
2006
Taira, Y.(National Fisheries Univ., Shimonoseki, Yamaguchi (Japan)) | ; Morimoto, E. | ; Nakamura, M.
This report deals with a discrimination method of fish species with neural networks. In the authors' previous works, the feature parameters about color information used as the inputs of neural network were set at the points of a fish image, and the experiment showed that such a setting method of color data has a poor discrimination performance. The objective of this report is to find an effective setting method of color data for fish discrimination. Firstly, the method in which the feature parameters were set not at the points but in the areas of a fish image was considered, and the effectiveness of this method was confirmed by the experiment. Secondly, the problem of reduction in the number of color parameters was addressed, and it was showed in the experiment that the number could be reduced from 27 to 6.
Показать больше [+] Меньше [-]