Define distance isn't changing calculated distance

Hey guys first post. I have been able to get to where I am with searches and a couple questions on the FB group but I decided it’s time to join the community as I need a little more help with a few things. Right now it is scaling my current model. 
I have a big model I am working on with about 2000 pics. UAV and ground. It required a quite a few control points to get them to tie together and decent alignment but it’s looking pretty good now. This is a lot I will be putting a modeled building on and so I want to scale the lot model. I used two control points between two fence posts I measured, defined the distance and ran alignment again but it doesn’t change to what it’s supposed to be. The defined distance is 5.2 meters. The actual distance before adding the constraint is something like 5.8. Once I add the constraint and run alignment it goes to something like 6.1. If I then change the value from 5.2 to anything else and update the calculated distance increases. So for example I changed the defined distance to 3 meters, updated, and then it said the calculated distance was 6.1 meters. Then I tried 7 meters and it changed to 6.6 meters. After that any value I enter only changes the calculated distance by a couple hundredths of a meter. 

I opened a small project I have been working on with a couple hundred photos, arbitrarily selected a couple of control points, added a defined distance, ran align again and the new component reported the calculated distance the same as the defined distance. Why isn’t it doing this in my bigger model? 

 

So in reading a little more if the camera such as a drone provides GPS the model will scale based off the GPS. 
Now my drone photos GPS points have always deviated significantly in many different ways. Now I did use two different drones so I have wondered if for example their altimeters were calibrated differently but regardless they skew considerably on x and y as well. I didn’t care much because I got everything aligned although I do wonder if this was why it was so hard to do so. But if it is going to screw with my scale this is definitely something I need to resolve.

hello and welcome, what version of RealityCapture are you using and how does it behave when you define more distance constraints at once?

Hi Summer,

 

The residule error lines in that screenshot aren’t that uncommon for UAV photos. Always a bit of disagreement between the original GPS position and what RC has figured out from photogrammetry. Especially if you used two different UAVs. Generally altitude is pretty hit and miss for GPS.

That aside.

Your added in tie points with scaling have to compete against all those GPS points. That is, the manual tie points with the set scale add are trying to influence the scale of the overall model but then so are those 2,000 photos all with GPS points.

So what you’re finding is RC is trying to find a solution that satisfies the condition of the manual tie points you set but in the end the 2000 GPS points have more weighting.

So you can either increase the “weighting” for the manual tie points or disable some or all of the GPS for the photos (can be done within RC). Either of those solutions will relax the amount of points influencing the scale of the model and hopefully let your manual points have a bit more influence and get a value closer to what you are inputting.