Stockpile Volume

[Claudito]

Hi everybody. I'm new to this world of point clouds! I'm having a big trouble to achieve a volume calculation with the "volume 2.5D" of my stockpile. I am trying to validate cloudcompare with a homemade experiment for later use in an industrial enviroment.

First of all i've had searched the forum with the keywords "stockpile volume" and i haven't found the answear to this. I've seen some similar issues but not exactly an answear to this.

I've tried two set ups with different materials.
To make the stockpiles i've used a precision scale a weighted water in a tea cup to know it's exact volume, then i filled up the cup with the materials and spread them in the floor to form the stockpile. I've made photogrammetry models with a very good results in the mesh generated.
i have modified the scale of this models to match a reference in the scene with known dimensions.

With all this done and checked i've segmented the cloud and tried to learn the volume with two methods.

The first one is explained in this video ... basically after segmentation i take the remaining part (the ground level), rasterize it with interpolation to fill the hole left by the segmentation. That leave me with a ground that can copy any slope in the terrain. I set this new raster ground as, exactly, ground level and compute with VOLUME 2.5D between the part cloud and the remaining rasterize cloud.

The second method is to take that segmented part with the stockpile and set a ground level by myself.

Anyway, the two methods with two different stockpiles it's giving me a volume 20-25% bigger than the real volume.
I know this must be something i'm doing wrong, i see the potential of this tool and i really want to understand it. I will post some photos of the model i'm talking about. If anyone could point me in the right direction for this i would appreciate it so much!!!

PD: i've tried to attach the cloudcompare file here but it says to me "invalid file extension" to the .bin file. Sorry for the bad english

foro1.png (460.35 KiB) Viewed 6685 times

[daniel]

Which 'step' did you use for the gridding in the 2.5D volume calculation tool? If it's too low, the approximation can be quite big.

Otherwise, don't hesitate to send the data by email (admin [at] cloudcompare.org) along with the true volume and also some reference dimensions if you have them visible in the clouds?

[Claudito]

Hello Daniel ! thanks for the answear i have used 0.05 as step! it looks fine to me.

I will send you the dataset now thorugh the "contact us" button in the bottom of the page. Is that ok?

[WargodHernandez]

just for curiosity, could you take model of the surface before the stock is added on top, then add the stock, rather than filling the gaps (since it appears to be a large portion of the base surface) and then align the edges of the surface together in the before/after adding the stockpile.
Also when comparing against your known artifact does the artifact have a "height" component? I want to validate you aren't calibrating against scales that are in plane with the ground surface only.

[daniel]

Yes. Hopefully it should lead you to the same email address: admin [at] cloudcompare.org (replace [at] by @)

[Claudito]

Daniel, I've just sent you the mail. Please tell me if it reaches you.

Thanks

[daniel]

I finally found the time to look at your data.

As the 'floor' was missing, I picked some points on the "gray" plate and then fit a plane through these points:

ground_extraction.JPG (137.7 KiB) Viewed 6431 times

Then I sampled a lot of points on both meshes (10M.) and used the 2.5D Volume Calculation tool:

Volume1.JPG (79.52 KiB) Viewed 6431 times

So one can indeed see that the the volume is 1200 cm^3 (while it should apparently be 984 m^3).

However, as the grid is pretty refined, I would say that the calculation is 'accurate' (from a numerical standpoint).

Therefore I looked at the scale. I checked the length of your "stickers" (I would say this is not the most accurate way to scale a photogrammetric cloud. You should have in your scene either a ruler on which you can make more precise measurements, or ideally a 3D object of known dimension such as a box, that you can best fit afterwards with ICP for instance).

I personally think your scale is a bit off (of roughly 5%). If the clouds are rescaled of -5%, then the volume is reduced to 1000 cm^3 (which is not that far from the target). Considering that the 'ground' is not that flat (at least in the data) and the various approximations, it wouldn't be that bad.

[Claudito]

Daniel first of all, and again, thanks for taking the time of analazing my data.

1) I would like to know wich tool did you use to pick the 10 points ( i think i've already used the "fit plane tool")?

2) i will try again the test and place in the scene a box of known dimensions.

I have a doubt about your conclusion of the scale. I had already notice that my scale wasn't that accurate, and this question is probably a fool one, but if the scale is wrong at about 5% how could this make a global error in the volume of aprox. 15%-20% ?

[daniel]

1) The Point List Picking tool

2) Yes that would be interesting

And regarding the scale, it's probably because if you scale all dimensions by 0.95, the volume becomes 0.95^3 = 0.85 times the original one.

[Claudito]

Daniel after i asked you i've adjust the scale of the dataset and create the plane from the point picking list.

The result it's way better, i'm still 10,9% away from the real volumen but getting closer.

To do this in the most accurrate way what do you recommend me apart from the "box reference" in the scene? Is there anything else i could do to get a best approximation, say in the 2-5% of error margin.