I've been searching the forum and read all the manuals avalible for Context Capture Master and the process of aerotriangulation and productions of orthophotos. But I'm still uncertain on how the process should be conducted in order to achive best result in shortest time. There are - in my perspective - too many options and settings to be altered without any proper explanation on how they affect the process and the final result.
So my question is, has anyone developed a standard process that could be described and applicated on a project like mine?
We have houndreds - or even thousands of photos shot with at simple consumer grade drone with a 12Mp camera from 80m height, 70/80% overlap on photos and it should all now bo georeferenced by survey grade control points - GCP:s collected with our GNSS-equipment. Our goal is to produce up to date orthophotos with 2,5 to 5cm resolution per pixel.
All our trial projects, so far, have come out with decent results, but that is after several trials and as it seems they may be successful or not, regardless of some of the settings we try. Unfortunately forrest parts won't be included in the photo process and house roofs come out with rounded corners. I can't get my head around things like the difference between "adjustment constrains" and "final rigid registration" although it is briefly described in the user manual. Also I find it strange when adding control points to the project and they show up all over the model - or atleast several at the time on single photos in the process of pointing out their position.
Any help is appreciated!
Anders Theodorsson, Ludvika Municipal GIS-office
Hello Scott
Well. I kind of got some answers on some of my questions. But I can't say I got an complete working order or final instructions on what settings to use. More like "try this and that" with the hope I will find my way forward anyway.
So I'm fiddling on and still have quite a few questions open but can at least make those orthophotos we need so desperately.
I might take a rainy day to sort my open issues out and list them in a new service request or what would be your suggestion?
When you suggest running with "heavy downsampling", do you mean reduced resolution, precision or number of photos? It might be an obvious answer to this that I'm not getting so forgive me if tha's the case!
I don't have answers to your questions about why photos with GCPs in them are being tossed out. except to reiterate what Oto said about them perhaps having the wrong coordinate associated with them. I've picked wrong GCPs before and been stumped as to why they didn't go together.
As for forest images being rejected, try flying higher to include some more recognizable objects in those areas.
For better roof corners and facades (and just generally) I take my images with a 70 degree gimbal angle. It does a better job with orthos as well as 3D models. A little more info on the edges helps crisp them up. This video also shows some techniques for improving corners on buildings as a post processing option.
I almost always just run defaults on all processes and it works. I assume that you started there too. I'm assuming that that means that the issue is with the GCPs or the capture.
As for downsampling, it is in the photos tab. I was suggesting downsampling for the AT so that you could run the process quickly to see if it works or not before you spend a lot of time running an AT that is bad. Then you can work to fix the issue before you waste a bunch of time. You can then remove the downsampling before you run the production. You can see from the attached image how much you can reduce the dataset. I've also done some tests downsampling the final productions as well. You can speed things up without a huge loss in quality with a little downsampling. I test this by selecting a small piece of the entire project in the Spatial Framework tab and running short productions at different downsampling levels to find the sweet spot.