ID: 1865
Presenting Author: Yohan Chang
Session: 665 - Planning Support Systems for Environmental Assessment and Urban Decision-Making
Status: pending
AI-based 3D spatial data cube integrates digital twins and machine learning to support climate-responsive environmental assessment and urban planning decision-making.
The increasing complexity of urban environments demands advanced planning support systems that can integrate multi-dimensional spatial data with artificial intelligence. This study presents a 3D Spatial Data Cube–based Machine Learning Framework developed under the National Spatial Information Advancement Project of Korea, aiming to enhance environmental assessment and urban decision-making.
The framework introduces a multi-layered voxel-based data cube that harmonizes heterogeneous spatial datasets (topography, buildings, mobility, environmental layers) into a consistent 3D reference structure. Machine learning modules—implemented using sparse convolutional neural networks via MinkowskiEngine—enable semantic segmentation, spatial classification, and scenario simulation across urban layers. The system adopts synthetic and labeled training data generation pipelines (e.g., STPLS3D-based point cloud labeling, Ground Truth creation, and synthetic augmentation) to address the inherent data sparsity of 3D grids.
Preliminary tests demonstrate the framework’s capability to perform 3D object recognition and environmental feature extraction, forming the foundation for a geo-intelligent planning support platform. The envisioned application extends to urban heat mapping, microclimate simulation, and infrastructure risk diagnostics, allowing planners to evaluate alternative policies in a digital-twin environment.
By embedding this AI-enabled spatial data cube within a national digital twin ecosystem, the study highlights new pathways for climate-responsive, data-driven, and participa
Yohan Chang is a research associate at KRIHS, specializing in GeoAI, digital twins, and AI-based spatial analytics for sustainable urban and environmental planning.
Coauthor 1: Jae Soen Son