2018/02/17 by Jeremy Castagno, Castagno, Jeremy D., Ella Atkins +1
Environmental Science · Engineering · #Remote Sensing and LiDAR Applications #Satellite Image Processing and Photogrammetry #Automated Road and Building Extraction
paper · pdf · doi:10.48550/arxiv.1802.06274
Geographic information systems (GIS) now provide accurate maps of terrain,\nroads, waterways, and building footprints and heights. Aircraft, particularly\nsmall unmanned aircraft systems, can exploit additional information such as\nbuilding roof structure to improve navigation accuracy and safety particularly\nin urban regions. This paper proposes a method to automatically label building\nroof shape types. Satellite imagery and LIDAR data from Witten, Germany are fed\nto convolutional neural networks (CNN) to extract salient feature vectors.\nSupervised training sets are automatically generated from pre-labeled buildings\ncontained in the OpenStreetMap database. Multiple CNN architectures are trained\nand tested, with the best performing networks providing a condensed feature set\nfor support vector machine and decision tree classifiers. Satellite and LIDAR\ndata fusion is shown to provide greater classification accuracy than through\nuse of either data type individually.\n