the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 23 Jan 2025
FACA v1 – Fully Automated Co-Alignment of UAV Point Clouds
Abstract. We introduce FACA – Fully Automated Co-Alignment, an open-source software designed to fully automate the workflow for co-aligning point clouds derived from Unmanned Aerial Vehicle (UAV) images using photogrammetry. We developed FACA to efficiently evaluate fieldwork with Unmanned Aerial Vehicles (UAVs) on landslides and coastal dynamics. The software applies to any research requiring comparative precise rapid multi-temporal point cloud generation from UAV imagery. UAVs are an essential element in most contemporary applied geosciences research toolkits. Typical products of UAV flights are point clouds created with photogrammetry, to measure objects and their change if multi-temporal data exists. Ground Control Points (GCPs) are considered the best method to increase the precision and accuracy of point clouds, but placing and measuring them is not always feasible during fieldwork. Co-alignment leads to the local precise alignment of multiple point clouds without GCPs. FACA uses Agisoft Metashape Pro and the Python standard library. The GPLv3 licensed FACA source code focuses on extendability, modifiability, and readability. FACA works interchangeably from the command line or a custom graphical user interface. We distribute the software with both usage and installation instructions. Three multi-temporal test datasets are available. We demonstrate the utility and versatility of FACA v1 with a multi-year and -region dataset acquired along Germany's Baltic Sea coast. FACA is in continuous open development.
- Preprint
(7099 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on gmd-2024-209', Becky Collins, 21 Feb 2025
General Comments
The paper provides an enhanced methodology for aligning UAV SfM photogrammetry derived point cloud data based on co-alignment techniques. The paper highlights many of the available methods for aligning UAV collected photogrammetry data to generate point clouds that can be used for change detection purposes across multi-temporal datasets. The presented technique represents an improvement in the implementation of co-alignment and provides a robust discussion of its implementation in two different field examples.
Specific Comments
The introduction of structure from motion would benefit from a couple of additional sentences of detail to provide more background on the technique and to give context for the subsequent work.
I think the introduction should contain quantitative critique of the existing methods. There is very little comparison between Direct and Indirect methods for georeferencing or the difference between the accuracy of a point cloud generated from GCP aligned data and that making use of a co-aligned or indirect method of alignment. For a newcomer to the field who may be reading this, it might be helpful for you to define the differences in accuracy between different existing techniques to highlight expected error. This could be done effectively with a simple table of typical errors of the various techniques.
Line 219 you mention supplementing the UAV collected data with manually taken images – would be good to know the reasoning for this and what decision making was taken in the field to decide when and where to do this.
Tables 2 and 3 could be combined together into one. Table 2 as it currently stands seems superfluous.
Section 3.3 you discuss the use of C2C distance for comparison of stable areas but earlier you highlighted the problem of C2C distance often being a function of point density. Some additional justification for that choice would be valuable here. Section 3.4 you present the use of the M3C2 method to compare distance to the first field campaign. Again, explanation of why this choice was made vs the already established C2C method would be valuable.
Frequent use of very high precision in quoting numbers of points and seconds of analysis. Consider whether this is needed as it can mask the magnitude of the differences between techniques. Perhaps consider quoting points to the nearest hundred or thousand and times to the nearest minute.
Table 11 it is clear that the filtering of original chunk and the exporting of the point clouds represent only a very small proportion of the overall time. Excluding them from this table would allow more focus to be given to the components that have a significant impact on the processing time.
The discussion section contains a number of very long paragraphs that could be broken down into smaller sections to better elucidate the points. Breaking the discussion more clearly into sections that discuss the erosion recorded and the accuracy/evaluation of the FACA method would be clearer, especially as this is a paper presenting a methodology and not a paper describing the erosion processes.
There could also be some attention paid to the concision of the discussion – for example, line 370-371 contain a short sentence telling of the uncertainty of tracing cliff tops and the next sentence then attributes variation in coastal retreat to these uncertainties. However this could be achieved by rewriting to something like – “The variation between yearly coastal retreat values for different co-alignment parameterizations is most likely attributable to uncertainties related to the process of hand tracing the cliff top as a measure of coastal erosion.”
More detail related to the statement made at line 367 regarding movement of the RTK base station is required. What were the indicators that led you to that conclusion?
Captions of figures 5 and 6 need additional information to make it clearer what they represent.
Technical Corrections
None identified.
Citation: https://doi.org/10.5194/gmd-2024-209-RC1 -
RC2: 'Comment on gmd-2024-209', Anonymous Referee #2, 08 May 2025
This article presents a useful fully automated co-alignment tool for UAV point clouds. The manuscript is well organised and easy to follow.
I do not have any major issues with the article, however I found the logical arguments at the start of the introduction difficult to follow. I think this could have clearer narrative structure and be explained a little better. Also regarding the introduction I was missing an overview of existing software or alternative options and what this software adds to that ecosystem.
The description in the results in section 3.5.3 is OK, but I’m unconvinced it needs to be so long if the paper aims to introduce the tool. If there are other useful findings and significant points from this analysis perhaps they can be signposted at the end of the introduction as a contribution of the paper.
Minor comments
Line 16 “Structure from Motion (SfM) is the process of…” This sentence hangs without any context or explanation. Can you rewrite the point being made here?
There is then a big jump in logic to discussing DEM differencing, which is confusing to me and I think the logical structure of the second paragraph could be better.
Line 56: Are other open source software available to accomplish similar tasks or do they all lack some functionality? Some mention or discussion of this would be useful before introducing your tool.
Figure 1 needs to be introduced in the text before being presented.
Around line 145: I was wondering if the description of what each parameter does would be better presented in table 1? Or could a supplementary table be produced with both the parameter, its corresponding values and a simple description of its purpose included. This is by no means a requirement but it would be good to know it's been considered.
Line 235: How representative is the location used to test the local precision? There is not much discussion of this point, but it seems like there is a rather limited set of objects and conditions with which to assess the alignment with the test data. Is it possible to strengthen the justification of the test sites and identify any limitations? I think these are mentied in 3.5.2 but if they are known a priori it would be good to be upfront.
Section 3.5.3 – there is a long description of the features in the imagery, but it's not always clear that this is necessary to present the modelling software. I would be tempted to shorten this but there is nothing wrong with the section if the authors can justify the inclusion.
Line 373 – I assume no uncertainty should be minimal or insignificant uncertainty.
Line 413: “The additional hour required with FACA defaults, compared to the parameters from Cook and Dietze (2019), to match and align the Wustrow study area is attributed to background usage of the workstation, not differences in parameterization.” Could this not be controlled for better? It seems an unnecessary lack of precision in the results.
Citation: https://doi.org/10.5194/gmd-2024-209-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 150 | 46 | 9 | 205 | 10 | 10 |
- HTML: 150
- PDF: 46
- XML: 9
- Total: 205
- BibTeX: 10
- EndNote: 10
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1