Independently Identifying Noise Clusters in 2D LiDAR Scanning with Clustering Algorithms

Light Detection and Ranging (LiDAR) refers to a range imaging method for distance objects based on the principle of laser ranging. LiDAR environmental mapping technology is often highly praised for its precise mapping information with intricate features for various detection or tracking based applications. The research proposes a novel method for independently identifying and filtering noise clusters in 2-Dimensional (2D) LiDAR scans based on 2 distinct clustering algorithms of K-Means and Density-Based Spatial Clustering of Applications with Noise (DBSCAN). Results show DBSCAN to be the better choice as it is more robust and resistance to noise and outliers in the dataset and is capable of identifying clusters of any shape making it more versatile. Furthermore, to address the issue of dead zones present in LiDAR scanning, an innovative solution based on interpolating the discontinuous spatial results of the LiDAR scanning result to further reconstruct a 3-Dimensional (3D) viewing model by stacking multiple copies of 2D LiDAR scanning results with varying elevation is demonstrated by the results of the study to be a viable economical alternative for 3D LiDAR mapping.