Improved progressive triangular irregular network densification filtering algorithm for airborne LiDAR data based on...

Area of Science:

• Geospatial science
• Remote sensing technology
• Computer vision

Background:

• Laser point cloud filtering is crucial for processing Light Detection and Ranging (LiDAR) data.
• The Progressive TIN Densification (PTD) algorithm is a widely used, robust method for filtering LiDAR data.
• PTD algorithm's performance relies on the initial Digital Terrain Model (DTM) quality and parameter tuning for diverse terrains.

Purpose of the Study:

• To introduce an improved PTD filtering algorithm, PTD-MSCN, designed for enhanced performance in complex terrains.
• To address limitations of the classic PTD algorithm related to DTM quality and parameter sensitivity.
• To improve the accuracy and robustness of LiDAR point cloud filtering.

Main Methods:

• Proposed the PTD-MSCN algorithm utilizing a multiscale cylindrical neighborhood.
• Employed the multiscale neighborhood to obtain and densify ground seed points for a high-quality DTM.
• Implemented an iterative process with decreasing neighborhood radius and distance threshold to identify ground points and remove object points.

Main Results:

• The PTD-MSCN algorithm was evaluated on 15 benchmark LiDAR datasets from the ISPRS.
• Experimental results showed a reduction in average total error from 5.31% to 3.32% with optimized parameters.
• The PTD-MSCN algorithm demonstrated superior accuracy and robustness compared to five other PTD filtering algorithms.

Conclusions:

• The PTD-MSCN algorithm effectively enhances LiDAR point cloud filtering, particularly in complex terrains.
• The multiscale cylindrical neighborhood approach leads to a higher quality DTM and more accurate ground point classification.
• PTD-MSCN offers a more accurate and robust solution for LiDAR data processing compared to existing PTD methods.