Semi-Supervised Learning for Land-Use Classification from Aerial Photograph Using Convolutional Neural Network

Abstract

In recent years, GIS (geographical information system), integrating various map information, has been utilized in a wide range of flelds. Land-use is useful information in GIS data, a basic data for developing and planning public works projects. However, detailed information is not available except for some urban areas. On the other hand, it is expected that detailed estimation can be performed by using machine learning, such as a convolutional neural network (CNN) from aerial photographs. However, to improve its accuracy, a large amount of labeled data is required. In this study, we consider generating efficiently labeled data from map symbols of GIS data as a means to efficiently increase the data. Further, we propose to perform semi-supervised learning using this method. We demonstrate the effectiveness of the proposed method in 6 classes of land-use classiflcation.