Problem

Airborne LiDAR point clouds contain both ground and non-ground returns. Ground filtering is the essential first step for DEM generation, building extraction, and vegetation analysis. Existing commercial solutions are expensive; existing open-source alternatives are often slow or struggle with complex terrain.

Approach

• Implement the progressive TIN densification algorithm, iteratively adding ground points to a triangulated irregular network based on distance and angle thresholds.
• Use CPU parallel processing to accelerate the iterative densification loop for large point clouds.
• Adopt an automatic block-based strategy that partitions data by spatial tiles, enabling processing of arbitrarily large datasets within constrained memory.

My Contribution

I independently designed and implemented the complete software from scratch, including the core algorithm, multi-threaded parallelization, block-based memory management, and command-line interface.

The software is released as freeware for non-commercial use and has been adopted by users in the photogrammetry community.

Outcome

The filter reliably separates ground and non-ground points across diverse terrain types including urban, forested, and mountainous areas. Block-based processing allows handling datasets exceeding hundreds of millions of points without excessive memory usage.

Download: Baidu Cloud (password: vezv)