Automatic LIDAR building segmentation based on DGCNN and euclidean clustering

Abstract

There has been growing demand for 3D modeling from earth observations, especially for purposes of urban and regional planning and management. The results of 3D observations has slowly become the primary source of data in terms of policy determination and infrastructure planning. In this research, we presented an automatic building segmentation method that directly uses LIDAR data. Previous works have utilized the CNN method to automatically segment buildings. However, the existing body of works have relied heavily on the conversion of LIDAR data into Digital Terrain Model (DTM), Digital Surface Model (DSM), or Digital Elevation Model (DEM) formats. Those formats required conversion of LIDAR data into raster images, which poses challenges to the evaluation of building volumes. In this paper, we collected LIDAR data with unmanned aerial vehicle and directly segmented buildings utilizing the said LIDAR data. We utilized a Dynamic Graph Convolutional Neural Network (DGCNN) algorithm to separate buildings and vegetation. We then utilized Euclidean Clustering to segment each building. We found that the combination of these methods are superior to prior works in the field, with accuracy up to 95.57% and an Intersection Over Union (IOU) score of 0.85.