Preparing to measure the 500-year-old Sumoto Castle using point clouds
I decided to create teaching materials for distance measurement using point clouds in Navisworks Freedom.
Since I'm at it, I want to use publicly available point clouds rather than sample data.
So, I looked for point cloud data that is available for free.
What I found was the Sumoto Castle Ruins 3D Point Cloud Data from Sumoto City, Hyogo Prefecture.
The year 2026 marks the 500th anniversary of the construction of Sumoto Castle.
This is quite interesting.
What kind of castle is Sumoto Castle?
Sumoto Castle is a mountain castle built on Mount Mikuma in the southeast of Awaji Island.
It is said to have been built by the Atagi clan in 1526.
Later, in 1585, Yasuharu Wakisaka entered the castle and developed the large-scale stone walls that remain today.
One of the features of Sumoto Castle is the 'climbing stone wall' that ascends the mountain slope vertically.
Even today, castle ruins such as stone walls, baileys, and the castle tower base remain.
And the point cloud data I found this time is not simply a scan of 'just the castle'.
It was quite extensive data that included the surrounding terrain from the foot of the mountain to the summit.
In other words,
Foot of the mountain
↓
Mountain path
↓
Slope
↓
Bailey
↓
Stone wall
↓
Sumoto Castle at the summit
The space itself is a point cloud.
This looks like it will be interesting as a teaching material.
I found the point cloud of Sumoto Castle
Sumoto City publishes 3D point cloud data of the Sumoto Castle ruins.
The data, measured using terrestrial laser scanners and Lidar SLAM, consists of a high-density point cloud with over 400 points/m² across the entire area.
Publicly available materials and data can be found here.
Sumoto Castle Ruins 3D Point Cloud Data
https://www.geospatial.jp/ckan/dataset/sumotojoseki
Sumoto Castle Ruins | Sumoto City Official Website
https://www.city.sumoto.lg.jp/site/tunagarumachi/16885.html
About the Sumoto Castle PR Video | Sumoto City Official Website
https://www.city.sumoto.lg.jp/soshiki/19/24182.html
A PR video is also available on YouTube.
In this video, vegetation is removed from the point cloud to showcase castle ruins such as stone walls.
First, I had to obtain the data.
This is where it got difficult.
The data for the Sumoto Castle ruins is published in segments covering a wide area.
I acquired the point cloud data in LAS format this time.
46 files.
In total, it was about 60GB.
Each file was about 300MB.
Just because the data is publicly available for free doesn't mean the teaching materials are ready with a single click.
In the end, the download alone took an entire day.
It seems that "free" does not necessarily mean "effortless."
Integrating with CloudCompare
The next day, I checked the 46 downloaded files in CloudCompare. Here, I was saved by one thing. The 46 LAS files shared the same base point from the start. In other words, I didn't need to align the positions for each file. If the base points had been different, simply merging them wouldn't have been enough. I would have needed to align the positional relationships between the point clouds, including not just moving them but also rotating them. It's not like CAD, where you can easily align things based on surfaces or edges. With 46 files, that would have been a very arduous task. This time, that step wasn't necessary. Since handling 46 files at once is too heavy, I merged them in batches of about 10 files at a time.
I integrated them step-by-step while saving the intermediate data in BIN format.
Ultimately,
1: Approximately 1.1 billion points
could be integrated.
Furthermore, the remaining 2 data was,
approximately 800 million points.
Combining 1 and 2 results in approximately 1.9 billion points.
Therefore, I tried merging the 1.9 billion points into a single PointCloud.
The result was,
Not enough memory!
With the 64GB of memory in my environment, integration as a single PointCloud was not possible.
However, I learned one thing here.
"Is it necessary to make everything into a single PointCloud?"
Trying it out in ReCap
So, I decided to export the data created in CloudCompare as LAS and try it in Autodesk ReCap Pro.
1 is,
approximately 1.1 billion points
approximately 27.9GB
in LAS format.
I loaded this into ReCap.
Processing took about 30 minutes.
While using a significant amount of CPU, the loading completed successfully.
Next, I added the second dataset, consisting of approximately 800 million points and about 15.8 GB, to the same ReCap project.
This took about 15 to 20 minutes.
During the process, there was a period where the progress bar seemed stuck around 50% for a while.
However, the processing itself had not stopped.
Ultimately, the addition of the second dataset was completed.
In other words,
approximately 1.1 billion points + approximately 800 million points
While I could not merge these point clouds, equivalent to about 1.9 billion points, into a single PointCloud using CloudCompare, I was able to import them into the same project in ReCap.
This was a very interesting part of this verification.
"Combining massive point clouds into one" and "handling multiple point clouds in a single project" are not the same thing.
I've come this far.
Currently, I am in a state where I can handle the wide-area point cloud of the Sumoto Castle ruins in ReCap.
Next, I will try to format this data so it can be handled in Navisworks Freedom.
Furthermore,
a full data version
a thinned-out, lightweight data version
I would like to create these and compare the actual operational feel and data sizes.
The ultimate goal is to create teaching materials that use point clouds to measure distances in Navisworks Freedom.
However, I don't know yet if Freedom can actually handle this data, which is equivalent to about 1.9 billion points.
That remains to be seen.
That is all for now.
Starting with a single point cloud dataset made available for free,
Find data
↓
Download 46 files
↓
Merge in CloudCompare
↓
Hit memory limits
↓
Try a different method in ReCap
↓
Import the equivalent of approximately 1.9 billion points into the project
I have reached this point.
Point clouds are not finished once you obtain the data.
Which software, which format, how much data, and how to handle it?
By actually working with it, those aspects start to become clearer little by little.
And finally, next is Navisworks Freedom.
Will I actually be able to measure this Sumoto Castle point cloud?
The rest will be in the next verification.
This time, I used the point cloud data released by Sumoto City. I think it is very grateful that such real data is made available for free.
Reference
Sumoto City 'Sumoto Castle Ruins'
G-Spatial Information Center 'Sumoto Castle Ruins 3D Point Cloud Data'
Sumoto City 'Sumoto Castle PR Video'
Autodesk ReCap Pro
CloudCompare