Reproduce “DMNR: Unsupervised De-noising of Point Clouds Corrupted by Airborne Particles” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Chu Chen, Yanqi Ma, Bingcheng Dong, Junjie Cao
Researchers from City University of Hong Kong and Dalian University of Technology developed DMNR and DMNR-H, unsupervised filter-based algorithms for de-noising LiDAR point clouds corrupted by airborne particles. These methods demonstrated superior performance, achieving an F1-score of 91.35 on real snow data, which surpassed previous unsupervised methods and was competitive with a supervised deep learning model, while preserving environmental details.
Key insights
- Noise points from airborne particles can be effectively distinguished by their typical low density, low intensity, and close range, in contrast to legitimate environmental structures.
- Dynamic, adaptive thresholds that consider a point's distance, intensity, and local density enable robust noise removal across varying adverse conditions and LiDAR sensor characteristics.
- Post-processing with clustering algorithms like HDBSCAN can refine de-noising results by recovering environmental points initially misclassified as noise, particularly in datasets with uneven density.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.