Technical gazette, Vol. 33 No. 5, 2026.
Original scientific paper
https://doi.org/10.17559/TV-20250908002968
Adaptive Weighted Fusion Denoising of LiDAR Sparse Point Clouds Based on Multi-View Geometric Consistency
Lin Zhang
; School of Electronic Information Engineering, Geely College, Chengdu City, Sichuan Province, China
*
Qian Wang
; School of Electronic Information Engineering, Geely College, Chengdu City, Sichuan Province, China
Ting Guo
; School of Electronic Information Engineering, Geely College, Chengdu City, Sichuan Province, China
* Corresponding author.
Abstract
This study proposes a Geometric-consistency-based Weighted Adaptive Fusion (GWAF) algorithm for denoising sparse LiDAR point clouds. The method integrates four key parameters, distance, angular continuity, ring number, and reflection intensity, into a unified geometric-weight system. This integration facilitates both adaptive neighborhood construction and dynamic thresholding. By establishing an intensity-radius mapping and applying multi-view consistency, the GWAF algorithm enhances boundary retention and weak target detection. Experimental results, using data from a RoboSense M1 LiDAR, demonstrate that the algorithm achieves a 2.4% improvement in denoising accuracy (0.8888) and a 15% reduction in false deletions. These performance gains surpass those of traditional radius and statistical filters, all while maintaining real-time efficiency.
Keywords
adaptive filtering; environment sensing; LiDAR; multi-view constraints; point cloud processing
Hrčak ID:
350421
URI
Publication date:
31.8.2026.
Visits: 0 *