PCA-Based Denoising Algorithm for Outdoor Lidar Point Cloud Data.

Area of Science:

• Computer Vision
• Geospatial Data Processing
• 3D Data Analysis

Background:

• Lidar point cloud data (PCD) is susceptible to plane noise in complex environments, hindering accurate 3D reconstruction.
• Existing noise removal techniques often lack efficiency or sufficient performance for complex scenes.

Purpose of the Study:

• To develop an efficient and effective filtering scheme for removing plane noise from lidar PCD.
• To improve the accuracy and speed of 3D reconstruction by enhancing lidar data quality.

Main Methods:

• Projection of 3D PCD onto a 2D plane for ground and wall data separation using grid statistics.
• Unsupervised segmentation of ground data via KD-tree and normal vector analysis.
• K nearest neighbor (KNN)-based segmentation with optimization for enhanced noise removal.

Main Results:

• The proposed grid PCA and ground splicing method effectively separates ground and wall data.
• The KNN-based segmentation significantly improves noise removal performance.
• The method demonstrates superior denoising performance and faster run speed compared to traditional approaches.

Conclusions:

• The developed filtering scheme offers a robust solution for lidar PCD noise reduction.
• This approach facilitates more accurate and efficient 3D reconstruction from noisy lidar data.