Cemetery UAV survey

[M81]

Hi all,

We are going to carry out a UAV survey of a cemetery where the clients wants, apart from the point cloud, to be able to see elevations on each gravestone and import all that data into a GIS system.

Is there a way to tag every gravestone with an elevation and then export all that in a table and bring into GIS? Would classification help in this case? We are talking about 20,000 gravestones of different shape and size.

Any thoughts would really be appreciated!

Regards,
Efstratios

[steves01x]

No help with the questions but are they wanting to be able to read each grave stone as well?

[JFlammVDC]

I'm guessing each of the gravestones are lying flat?

[Martin_au]

I think the solution depends on ‘why’ they want elevations for each grave.

[gordonired]

I've been scratching my head regarding this question today. The really slow way that I would tackle this is: Cut out all the "ground" points using the CSF plugin in cloudcomparehttps://www.cloudcompare.org/doc/wiki/i ... 8plugin%29 so that you'll be left with only headstones.

I'd then have to go in and crop each and every headstone one by one into individual point clouds, and then export them as .txt, import into excel and sort the Z/height column by the largest number. It's a really really slow/cumbersome solution, but I haven't seen anything else posted so there ya go.

I'm curious to see if there is a better way to do this, preferably using free tools? And I know it only tackles one part of the original question, best I can do.

[MikeDailey]

More information about this project would be helpful.

If I were doing this, I would first decide if I needed to use lidar. It may be easier to see the stones with photogrammetry but if the yard is unkept or contains many trees, lidar is your choice.

What type of stones are these? If they're flat, ground markers, you may have to fly fairly low/slow to pick them out depending on your sensor. I see a bunch of velodyne/hesai sensors out there these days. They're great entry lidar options but the intensity returns and precision are greatly lacking. These "puck" type sensors are going to have a harder time picking up small flat targets or thin, standing headstones.

What software do you have access to? Hopefully you have some type of classification workflow for this. We use topodot, so I would classify a layer as all points from .5'-3' above ground. This would allow quick visualization of any standing stones and then using an extraction tool to pull a point for each.

Lastly do you need surface points at each stone's location or do you need the height of each stone? Stone height is going to vary in accuracy depending on your use of photogrammetry or lidar. Depending on which elevation you needed, you could simply extract the highest or lowest point in the defined extraction box if you have access to topodot.

[smacl]

I've been scratching my head regarding this question today. The really slow way that I would tackle this is: Cut out all the "ground" points using the CSF plugin in cloudcomparehttps://www.cloudcompare.org/doc/wiki/i ... 8plugin%29 so that you'll be left with only headstones.

I'd then have to go in and crop each and every headstone one by one into individual point clouds, and then export them as .txt, import into excel and sort the Z/height column by the largest number. It's a really really slow/cumbersome solution, but I haven't seen anything else posted so there ya go.

I'm curious to see if there is a better way to do this, preferably using free tools? And I know it only tackles one part of the original question, best I can do.

Was thinking something similar and we have tools in SCC to automatically vectorise and tag results in this type of scenario, see this video that does something similar for road surface analysis. Would work fine for upright gravestones but flat gravestones would likely be part of the ground. In this case they'd need to be re-classified somehow in order to be automatically vectorized. I'd recommend looking at Florent Poux's work on semanitc segmentation for this, see https://medium.com/towards-data-science ... 32618ca5df

So much comes down to the quality of the data, interference factors such as scrub on the ground, and the discernible geometry of the gravestones in relation to the ground.

[Martin_au]

I've been scratching my head regarding this question today. The really slow way that I would tackle this is: Cut out all the "ground" points using the CSF plugin in cloudcomparehttps://www.cloudcompare.org/doc/wiki/i ... 8plugin%29 so that you'll be left with only headstones.

I'd then have to go in and crop each and every headstone one by one into individual point clouds, and then export them as .txt, import into excel and sort the Z/height column by the largest number. It's a really really slow/cumbersome solution, but I haven't seen anything else posted so there ya go.

I'm curious to see if there is a better way to do this, preferably using free tools? And I know it only tackles one part of the original question, best I can do.

Cloud compare, csf or rasterize to create and chop out ground surface. Then rasterise to choose the highest point every ~1m or so to grab the highest point of each gravestone. This depends on what the purpose of the project is though.