This study proposed a novel approach for autonomously monitoring and analyzing pond bottom environments in kuruma shrimp aquaculture. An autonomous boat equipped with an underwater camera was developed to capture bottom footage, and the collected images were analyzed using a convolutional neural network (CNN) to automatically classify environmental conditions such as clean gravel, residual feed, white fungal growth, and sludge-like sediment. In addition, a blind annotation approach was adopted to objectively verify and evaluate the validity of the CNN classification results. These findings demonstrated that the proposed method can quantitatively and efficiently identify degraded bottom areas, providing a practical and labor-saving technology that contributes to improved productivity in aquaculture. The study was highly evaluated at the conference and received the Best Presentation Award. This study was conducted as a collaborative research project with Okazaki Co., Ltd., Prof. Hiroki Irie and Prof. Kiyoteru Hayama of the National Institute of Technology, Kumamoto College, and Prof. Takashi Okayasu of the Faculty of Agriculture, Kyushu University.
