3D Scan compared to Leica RTC360

[Ascent]

Hey hey all, So wanted to open up some discussion about the real accuracy of 3D Scan software.
What I have found is that we scanned a small lot with a storage shed on it to get some measurements with the Skydio 2+ enterprise and scanned the shed with a Leica RTC360 terrestrial laser scanner.
We use the RTC a lot and I knew the Skydio was not going to be as accurate but wanted to see how close it could get. I mean reading what the company states it is or could be survey grade in the cm range.
Well the shed is 31' x 14.5' and in that short distance there is an 7.7" east, 7.2" north and 6.3" west discrepancy in wall positions.
Now grant it we did not use GSPs (Which we will try next) we wanted to see what it did on its own. I am hoping that the putting it on control will fix this but wanted to see what everyone else was experiencing and just start up a discussion on best practice's.

I processed on both WebODM and Pix4d Mapper and both are virtually exact. The average GSD is 0.36cm/0.14" per Pix4D... That's 3.6mm So at an average of 3x's that is 10.8mm so basically 1cm supposed accuracy if I am doing my math right haaa haa

We use a lot of terrestrial, mobile and drones (other than the Skydio) to do our everyday scanning and know the tricks of the trade to get accuracy with their data but wanted to implement the Skydio of tight places and interiors. If it isn't more accurate than that it is basically only good for observation pic's. Far from what I was sold on.

So anyone out there up to a conversation to explore?

I just want to see if anyone has figured a way to get this tighter or has any ideas on processing

 

[LyleCDG]

Wow.. crickets for over a year. lol. I've been doing photogrammetry with drones for many years. It is unlikely to ever get accurate measurements without some survey Ground Control Points (GCPs). A rule of thumb is to use 5 GCPs for every battery used during flight. I try to use a minimum of 5 targets regardless of size of area. Typically I use more than five. We get very good, (1/10th foot accuracy) when using GCPs. We frequently do laser scanning AND a drone flight for some jobs. I can use data from the laser scan as 'targets' when processing the drone flight in Pix4D or Autodesk ReCap Photo with equally good results. But, no, I never use a point cloud for design unless we've processed it with GCP targets.

 

[Airwolf]

Using 3D Scan Map capture on the Skydio X10 and I am not getting a proper model when the scan is completed, and the images are processed? Any ideas? I am scanning a post wildfire area with a 60% slope on average. Area is about 20 acres.

Since I can't upload a tiff imager here I have attached a screen shot of the tiff imager the Skydio generates in the controller.

 

[LyleCDG]

If the slope is 60%, is your overlap consistent? If flying at a constant altitude, the overlap at the top of the slope will be less that the bottom of the slope. I'm not yet familiar enough with Scan3D to know if you can fly at a constant AGL altitude or not. Also, are you using ground control points to get a properly scaled map? All I saw in image you included is a small, long speck on a large white area. If that's what you're getting, do you perhaps have a couple photos in there with coordinates a couple miles away from your 20 acre site? Those are my thoughts.

 

[Airwolf]

If the slope is 60%, is your overlap consistent? If flying at a constant altitude, the overlap at the top of the slope will be less that the bottom of the slope. I'm not yet familiar enough with Scan3D to know if you can fly at a constant AGL altitude or not. Also, are you using ground control points to get a properly scaled map? All I saw in image you included is a small, long speck on a large white area. If that's what you're getting, do you perhaps have a couple photos in there with coordinates a couple miles away from your 20 acre site? Those are my thoughts.

Interesting points that got me thinking. I'm going to have to find a computer that has Agisoft Photoscan so I can manually stich the photos together to see if what you are alluding to might be the issue. I don't have authorization to install any other software to process the photos. Not Dronedeply, not any Pix4D, not Agisoft Photoscan except where it is already installed. So, this on-board option is our only option.

We did not have control points. Most of our work will never have control points because our terrain is forest land in Oregon on very steep and rugged terrain, the whole reason for the drone. All the photos are within a 20 to 30 acre plot. The X10 was never out of line of site.

The image I posted in my reply is what the drone produced when it processed the model. The slope is indeed between 50 and 70%. And overlap is 60. We set our flight elevation at 220 feet above launch point. Mapping area was all downhill. Lauch point was highest point. So, if it can only do a scan on a flat surface than it is completely useless for us. For $3000 a year it was promised to process the model from the images on board the X10 and output a 3D model. And what exactly is the difference between 3D Surface and Mapper? I used the Mapper. The whole point is to be able to produce a 3D model without having to use external software.

If what you are suggesting by your thoughts than the X10 is not good for our work. We were successful in producing models with the 3DR Solo and the Parrot ANAFI (consumer version) back in the day and didn't have any issue with slopes and overlap.