Automatic Volume Calculation and Mapping of Construction and Demolition Debris Using Drones, Deep Learning, and GIS

Abstract

This paper presents a time- and cost-efficient method for the management of construction and demolition (C&D) debris at construction sites, demolition jobsites, and illegal C&D waste dumping sites. The developed method integrates various drone, deep learning, and geographic information system (GIS) technologies, including C&D debris drone scanning, 3D reconstruction with structure from motion (SfM), image segmentation with fully convolutional network (FCN), and C&D debris information management with georeferenced 2D and 3D as-built. Experiments and parameter analysis led us to conclude that (1) drone photogrammetry using top- and side-view images is effective in the 3D reconstruction of C&D debris (stockpiles); (2) FCNs are effective in C&D debris extraction with point cloud-generated RGB orthoimages with a high intersection over union (IoU) value of 0.9 for concrete debris; and (3) using FCN-generated pixelwise label images, point cloud-converted elevation data for projected area, and volume measurements of C&D debris is both robust and accurate. The developed automatic method provides quantitative and geographic information to support city governments in intelligent information management of C&D debris.