Semantic-Based Building Extraction from LiDAR Point Clouds Using Contexts and Optimization in Complex Environment.

Area of Science:

• Geospatial Science
• Computer Vision
• Remote Sensing

Background:

• Building extraction from LiDAR point clouds is crucial but challenging due to complex structures and occlusions.
• Existing methods struggle with incomplete data and similar features across categories in urban and campus scenes.

Purpose of the Study:

• To present a versatile, hierarchical, semantic-based method for accurate building extraction from LiDAR data.
• To address challenges in complex environments like urban and campus scenes.

Main Methods:

• Preprocessing LiDAR data, including ground point removal and super-point establishment.
• Semantic labeling of raw LiDAR data and feature engineering focused on building descriptors.
• Employing a Markov Random Field (MRF) optimization model for postprocessing and refining segmentation results.

Main Results:

• The proposed method successfully extracts buildings from diverse LiDAR point cloud datasets.
• Experimental results demonstrate superior performance compared to state-of-the-art techniques.
• The approach shows feasibility and effectiveness across multiple complex environments.

Conclusions:

• The developed semantic-based method with MRF optimization is a robust solution for building extraction from LiDAR.
• This technique enhances the accuracy and reliability of building identification in challenging geospatial data.
• The study validates the method's effectiveness in varied urban and campus settings.