Authors: Wenwu Tang*, University of North Carolina at Charlotte, Shen-en Chen, University of North Carolina at Charlotte, Craig Allan, University of North Carolina at Charlotte, John Diemer, University of North Carolina at Charlotte, Matthew S Lauffer, North Carolina Department of Transportation
Topics: Cyberinfrastructure, Spatial Analysis & Modeling, Geographic Information Science and Systems
Keywords: Deep learning, Point cloud data, Hydraulic structures, 3D GIS, cyberinfrastructure
Session Type: Paper
Start / End Time: 9:35 AM / 10:50 AM
Room: Plaza Court 5, Sheraton, Concourse Level
Presentation File: No File Uploaded
The classification of point cloud data collected from LiDAR or sonar technologies plays an important role in the detection and measurement of 3D hydraulic structures such as bridge and culverts. However, this point cloud classification poses a significant big data challenge because of the volume and complexity of the point cloud data involved. In this study, we present a deep learning-based spatially explicit modeling framework for the automated point cloud classification of 3D hydraulic structures, referred to as DeepHyd. Deep learning, as a cutting-edge artificial intelligence techniques, has drawn increasing attention for its applicability in domain-specific studies. The DeepHyd framework is designed to classify and evaluate point cloud data for the detection of hydraulic structures based on deep learning algorithms. We trained the deep learning algorithm using cyberinfrastructure-enabled high-performance computing resources, represented by many-core graphics processing units (GPUs). This deep learning-based spatially explicit modeling framework is of great help to 3D geospatial studies, including transportation, hydraulics and hydrology, and watershed assessment.