Rockfall Detection from Terrestrial LiDAR Point Clouds by Using DBSCAN with ClutterRemoval Based on Grid Density

EasyChair Preprint no. 4637

Abstract

This paper proposes a simple method for rockfall detection from terrestrial LiDAR point clouds. The method consists of four steps: registration, subtraction, clutter removal, and spatial clustering. The paper contributes a straightforward method for clutter removal based on grid density, which is computational complexity inexpensive compared to the standard method based on nearest neighbor distance. Experimental results show that both are comparable in terms of identifying rockfall events. The proposed method can detect 21 events from 27 events from our simulations, and a conventional method can detect 23 events. The false-positive events of the proposed and conventional methods are 1 and15, respectively. In contrast, for 52,000 points, the proposed method is about 16 times faster. Also, this paper suggests a simple means to estimate the parameters used in the spatial clustering algorithm.

Keyphrases: Clutter removal, DBSCAN, Rockfall detection, Terrestrial LiDAR point cloud

@Booklet{EasyChair:4637,
  author = {Pitisit Dillon and Jessada Karnjana and Pakinee Aimmanee},
  title = {Rockfall Detection from Terrestrial LiDAR Point Clouds by Using DBSCAN with ClutterRemoval Based on Grid Density},
  howpublished = {EasyChair Preprint no. 4637},

  year = {EasyChair, 2020}}