Abstract
In an industry-government-academia collaboration for fixed asset tax evaluation work in the local government, we constructed a solar panel detection system from aerial photographs by utilizing large-scale land data and artificial intelligence technology owned by the local government. Currently, fixed asset tax is an essential financial source in municipalities nationwide. The municipal staff spends a lot of time and effort conducting field surveys to make appropriate evaluations and are required to utilize information and communication technology and data. Therefore, we have developed a solar panel detection system that trains object detection through deep learning. This artificial intelligence technology utilizes fixed asset information such as aerial photography images, land lot number map shapefiles, and taxable area data owned by local governments to develop the object detection function. In addition, we verified the detection system through a demonstration experiment in a local government through industry-government-academia collaboration. Then, we improved the system from the feedback of the demonstration experiment to practical use. As a result, we could confirm this system's effectiveness for solar panel detection and operational efficiency. In addition, we verify a series of efforts that utilize artificial intelligence technology and data for local government works and aim to activate data utilization in local governments.
