They did not accept our work

Hi. The customer and our dialogue illustrated with a picture. I’d like a DD. opinion of a staff member. Thanks.

Semilab measured the width of the south and east facades of the “k” building for us, but unfortunately its size does not match the dimensions of the outer point cloud.

  • Customer measurement; South side: 62.30, East side: 74.30
  • Point Cloud: South: 61.73, East: 73.64
    (- Our model: South page: 62.10, East page: 74.10) is given in meters.

We noticed the following today.

Below we see the west façade. There is theoretically a height difference of 15-20 cm between the points at the two ends, which are theoretically horizontal. The same is confirmed on the east façade. north, south horizontal.

It’s like the vertical isn’t vertical either.

Isn’t this point cloud skewed with this 15-20 cm?



The building is built on sloping terrain, there is a slope in the north-east direction (towards the south). The fact that the building will tilt a little towards the side below will sink in 100 years, it is almost natural, the chimney may tilt, but we will still check all this, take it for granted for the time being.

I would give a small chance for a pillar-framed building to tip evenly over 70m without cracking. But of course it can.

Gábor was outside, measuring the deviation from the horizontal plane from a stand at the top of the building, on the east and west sides at the top of the attic wall, and next to the building on the west side at ground level.
It became true, the drone point cloud model was indeed skewed by an average of 0.28 ° compared to reality. This has now been corrected and the point cloud has also been optimized for faster handling in Archicad with minimal loss. I dug myself into CloudCompare a bit and found good features, one of which saves unique, redundant points away from the central point population from the point cloud, reducing the number of points from 54M to 52M. The other is the “Subsample” function, where a minimal point cloud thinning has already resulted in an 18M point point cloud.
In the share link in the KulsoFelm \ 18Mpoint folder you will find the point cloud improved above and the Archicad v24 PLA file where I stared at the floor plan and sections.
There were many lessons to be learned from this drone survey, the model, which looks good from afar, is no longer as good when meticulously. The corrections made will improve the situation somewhat, but I still recommend that you adjust the model to the points measured by the laser scanner in the event of a collision rather than an external survey.
Sorry for the inconvenience!

1 month of preparation and $ 800 thrown out the window because of the many permits. I don’t think there will be a perfect understanding of correspondence, but the point can be read from it. Please tell me what I made wrong as a drone? I will send the finished 3D model and the route of my work. Thank you very much in advance.

1 Like

So basically they are wanting a building asbuilt where they are expecting the real world condition to perfectly match a plan and plumb vertically. Without RTK/PPK or at least GCP’s I think your results are about on par with what should be expected, especially when trying to measure a point cloud. I can guarantee you that chimney stack is leaning as shown. Every one of those that I have seen more than 20 years old is doing that. I am guessing this is some kind of renovation? If they are trying to hold it to architectural tolerances you will need to do RTK/PPK or a terrestrial laser scanner.

I don’t know how they field measured but I will say that when measuring in the field a lot of people forget about surface (slope) vs horizontal distance and shooting offsets to corners and tape measuring. Trying to shoot a corner directly or measure off a sparse point cloud (basically what drone vs ground scan point cloud) is bad form for renovations.

1 Like