RESEARCH ON FEATURE MAP IN ZOSTERA MARINA FIELD IMAGE DISCRIMINATION BY CNN

Abstract

 The eelgrass beds are important for coastal ecosystems, and there are many researches on monitoring. The authors are engaged in research to analyze the area of eelgrass beds from aerial photographs. The authors (2018) have shown that CNN can learn eelgrass field images and perform image discrimination with 94% accuracy. CNN is an algorithm that learns outline features and performs image discrimination in general. However, the eelgrass field image has few outline features, and the examination about the fact that the identification accuracy of the eelgrass field was high was not enough. In this study, we focused on pattern of feature map and verified discrimination.