APPLICABILITY OF AN IMAGE AI MODEL FOR CONFIRMING THE OPEN-CLOSE STATUS OF MARITIME LAND GATES

Abstract

 Due to giant typhoons in recent years, the development of a land gate open-close status confirming system is urgently required in order to alleviate storm surge damage, and to take immediate and wide-ranging evacuation activities and emergency measures. In this study, an image AI model using deep learning was constructed to confirm the open-close status of land gates installed as part of maritime embankments or levees, and its applicability was examined. Size and RGB of the images were examined, and the conditions of the image AI were assessed in order to develop the model for confirming land gate open-close status. As a result, the quality of the model was confirmed with 98% correct status designations. Examination of the image AI model using Grad-cam revealed that the status confirmation was judged from the land gates edges.