An Algorithm to Identify and Localize Suitable Dock Locations from 3-D LiDAR Scans
Abstract
Unmanned vehicles have established an important place in the modern battlefield. They play a key role in intelligence and surveillance while not putting human lives in harm's way. A necessary enabling technology is the ability to automatically identify Launch and Recovery sites from sensor data. This project focuses on the identification of suitable docking sites from three dimensional Light Detection and Ranging (LiDAR) scans. A LiDAR sensor is a sensor that collects range images from a rotating array of vertically aligned lasers. Our solution leverages open source C++ code from Point Cloud Library -- "a standalone, large scale, open project for 2D/3D image and point cloud processing." First the Random Sample Consensus (RANSAC) algorithm is used to isolate horizontal planar surfaces that may belong to the dock. After removing planar dock points, Euclidean Cluster Recognition is used to isolate point clusters that are potential vertical pilings. Bayes' Theorem is used to compute the probability that each cluster matches the characteristics of piling. For each candidate piling, the origin of that cluster is compared to the location of the target dock's planar surface. The dock can be identified by the relation of the pilings location to the dock's planar surface. The final output of the algorithm will be a sub-set of points, isolated from the original cloud, that are hypothesized to correspond to the dock.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 10, 2013
- Accession Number
- ADA581853
Entities
People
- Mitchell R. Graves
Organizations
- United States Naval Academy