Parallel Processing Method for Airborne Laser Scanning Data Using a PC Cluster and a Virtual Grid.

Area of Science:

• Geospatial Science
• Computer Science
• Remote Sensing

Background:

• Airborne laser scanning (ALS) generates vast datasets requiring efficient processing.
• Sequential processing methods can be time-consuming for large-scale ALS data.
• Digital Surface Models (DSM) and Digital Terrain Models (DTM) are crucial geospatial products.

Purpose of the Study:

• To develop and evaluate a parallel processing method for fast ALS data analysis.
• To enable the efficient generation of DSM and DTM from large ALS point clouds.
• To compare the performance and accuracy of parallel versus sequential processing.

Main Methods:

• A parallel processing approach using a PC cluster and virtual grid.
• Inverse Distance Weighting (IDW) interpolation for DSM creation.
• Local minimum filtering for DTM generation from DSM.
• Controlled handling of boundary data and interpolation centers across nodes.

Main Results:

• Demonstrated significant speedup and efficiency gains with parallel processing.
• Confirmed that parallel processing performance improves with increased computational overhead, processors, and data size.
• Verified that the parallel algorithm operates in linear time.
• Ensured identical DSM and DTM outputs compared to sequential processing.

Conclusions:

• The proposed parallel processing method is effective for handling large ALS datasets.
• This approach offers a scalable and efficient solution for generating geospatial models.
• Parallel processing is a viable and accurate alternative to sequential methods for ALS data.