Real-world comparisons between target-based and targetless point-cloud registration in FARO Scene, Trimble RealWorks and Autodesk Recap

Cox, Raymond (2015) Real-world comparisons between target-based and targetless point-cloud registration in FARO Scene, Trimble RealWorks and Autodesk Recap. [USQ Project]

[img]
Preview
Text
Cox_R_Zhang.pdf

Download (6Mb) | Preview

Abstract

With recent improvements to point cloud processing software, encouragement has been given by software manufacturers and retailers to register individual point clouds without isolated targets. Manufacturer test trials and online tutorials show examples of targetless point cloud registration in action. However, it is believed by the author that these examples, using only a handful of scans in beneficial scanning environments, do not necessarily take into account varied, real-world survey conditions that often take multiple scans to effectively measure.

This research project seeks to vigorously test and compare the accuracy of point-cloud registration between target-based and targetless processing methodologies across three common point-cloud software packages—FARO Scene, Trimble Realworks and Autodesk Recap. The project is largely field-work based and experimental in nature using six real-world survey scanning scenarios:
1) a heritage facade survey,
2) a quarry plant / conveyor system survey,
3) an indoor floorplan survey,
4) a tunnel survey,
5) a commercial roof survey, and
6) a wharf monitor survey

In each scanning scenario, target-based and targetless point-cloud registration methods have been compared and evaluated as to the accuracy of targetless registration for data extraction.

A control survey was undertaken for each scenario using traditional total-station measurements of key scanning targets within the scan area to form a base datum. The scanning data was then processed through each software package with the artificial targets isolated and point clouds registered as normal. The relative points were compared to the control survey and found to align well both horizontally and vertically. The identical (pre-edited) raw scan FLS files were then re-processed through each software package. The artificial targets within the scans (minus the isolated control stations that needed to remain for comparisons) were cut away and a targetless registration was completed for all survey areas utilizing as far as possible the software’s default targetless registration settings. The results were compared to the control survey and it was found that, for most scenarios, the control stations generally aligned horizontally within survey accuracies but vertically showed a lack of accuracy required for the precise nature of terrestrial laser scanning (TLS).

By testing across a good cross-section of survey scanning scenarios, it can be demonstrated that targetless point cloud registration has its limitations in accurately portraying certain real-world conditions especially in regard to vertical displacement. Though possibly suited to some surveying scenarios, vigorous quality control and traditional check measurements should be used to support user confidence of the data obtained.

Further research into developing better targetless registration algorithms that seek to minimize cloud to cloud vertical distortion would be beneficial to targetless software development/enhancement.


Statistics for USQ ePrint 29195
Statistics for this ePrint Item
Item Type: USQ Project
Item Status: Live Archive
Additional Information: Bachelor of Spatial Science project
Faculty / Department / School: Current - Faculty of Health, Engineering and Sciences - School of Civil Engineering and Surveying
Supervisors: Zhang, Zhenyu
Date Deposited: 01 Jun 2016 02:08
Last Modified: 06 Jun 2016 01:09
Uncontrolled Keywords: Point cloud, Registration, Target-less, Laser scanning, TLS, FARO Scene, Trimble Realworks, Autodesk Recap
Fields of Research : 09 Engineering > 0909 Geomatic Engineering > 090906 Surveying (incl. Hydrographic Surveying)
URI: http://eprints.usq.edu.au/id/eprint/29195

Actions (login required)

View Item Archive Repository Staff Only