Development of an Automated Settlement Mapping System using High-Resolution Satellite Images and Deep Learning

TitleDevelopment of an Automated Settlement Mapping System using High-Resolution Satellite Images and Deep Learning
Publication TypeConference Paper
Year of Publication2019
AuthorsMiyazaki, H, Ohira, W, Kaneko, S, Shibasaki, R
Conference NameThe 60th Annual Meeting for the Japanese Society of Tropical Medicine
Date Published11/2019
Conference LocationOkinawa, Japan
Abstract

We developed an automated system of building with the use of high-resolution satellite images and deep learning, comprising a geospatial data management system, an image data processing system, and a quality control system. The system development has achieved the component of the geospatial data management and image data processing, and performed building mapping of some large extents, while the development of quality control systems is ongoing. Because we developed the system with open-source and web-based software, anyone can participate in the preparation of training data just only with a computer and the Internet. The system is expected to be a platform for the large-scale mapping of buildings and other ground objects with international collaborations of local partners. The building maps developed by this system are expected to be a basis of analyzing demography with possible risks and impacts of communicable diseases.

Refereed DesignationNon-Refereed