[Building Change Detection Based on Multi-Level Rules Classification with Airborne LiDAR Data and Aerial Images]

Area of Science:

• Geospatial analysis
• Remote sensing
• Computer vision

Context:

• Accurate building change detection is crucial for urban planning and monitoring.
• Integrating heterogeneous data sources like LiDAR and aerial imagery presents unique challenges.
• Existing methods often struggle with the disparity between point cloud and image data.

Purpose:

• To develop a robust building change detection method that effectively fuses LiDAR point cloud and aerial image data.
• To introduce a multi-level rules classification algorithm and a morphological post-processing technique for enhanced accuracy.
• To provide a complete processing workflow applicable to real-world scenarios.

Summary:

• A novel building change detection method is proposed, combining LiDAR point clouds and aerial imagery through a multi-level rules classification algorithm.
• A morphological post-processing step with an area threshold refines the detection results.
• The method was evaluated using data from Changchun City, China, demonstrating high performance with a Kappa index of 0.90 and correctness of 0.87.

Impact:

• The proposed method significantly outperforms the object-oriented Support Vector Machine (SVM) classification approach.
• Achieves ideal performance metrics, indicating its effectiveness for practical building change detection.
• Offers a valuable tool for urban management, disaster assessment, and infrastructure monitoring.