Generating Seamless Three-Dimensional Maps by Integrating Low-Cost Unmanned Aerial Vehicle Imagery and Mobile Mapping...

Area of Science:

• Geomatics Engineering
• Computer Vision
• Urban Informatics

Background:

• Traditional 3D urban mapping faces challenges with data occlusions and limited detail from single sources.
• Integrating ground-based and aerial data offers complementary strengths for comprehensive urban modeling.

Purpose of the Study:

• To develop a novel framework for generating seamless, high-fidelity 3D urban maps by fusing mobile mapping system (MMS) and unmanned aerial vehicle (UAV) data.
• To overcome limitations of single-source mapping, such as aerial view occlusions and ground-level vertical detail deficiencies.

Main Methods:

• Utilized cloth simulation filtering for ground point extraction from MMS data.
• Employed deep learning (U²-Net) for feature extraction from UAV imagery.
• Applied inverse perspective mapping and LightGlue for cross-view data alignment and integration into a Structure from Motion pipeline.

Main Results:

• Achieved significant accuracy improvements in 3D urban models through fused data.
• Demonstrated a root mean square error of 0.131 m in validation datasets.
• Successfully generated seamless, high-fidelity 3D urban maps.

Conclusions:

• The proposed framework offers a flexible and scalable solution for enhanced 3D urban mapping.
• This approach improves geospatial analysis, infrastructure monitoring, and urban planning capabilities.
• The fusion of MMS and UAV data provides superior results compared to single-source methods.