Hi All,
How accurate are the Geo-location terrains? I have had a topological survey done of a site and downloaded the Hi-res information from Digital Globe. The picture shows the downloaded information and the contours created by the survey. I wasn’t expecting them to be an interference fit but at worst there is a 3m difference.
Was the survey done on the ground? The terrain available via Digital Globe isn’t created from on the ground surveys but from aerial LIDAR data. I expect it’s also smoothed a bit from the original data so I wouldn’t expect the precision you get from a survey of the site. If I were you I would use the texture image you got with the geolocation data to paint the terrain generated from the survey and discard the DG terrain.
Thanks Dave,
That is what I expected. It makes it pretty much a waste of time for what I need other than to set the sun position.
If you don’t need the imagery either, just manually enter the lat/long in Model Info>Geo-location.
It would be nicer if the geo-location information was accurate. It’s a nice picture and that’s it. Very disappointing.
I expect higher res data would come at a higher price not only financially but in an exponential increase in entity count which would cause performance problems and delays. They’d need to get surveyors on the ground and that doesn’t come quickly or for free.
Here is a test you can do:
Add Location, with something distinct on the + symbol.
Go to maps.google.com, and right-click on the exact same spot, and select the coordinates you are given in the in the first menu item. That will be copied to your clipboard.
In SketchUp, choose Add Location, Add Imagery, and paste in the coordinates.
Look to see what is located where the + is.
For me, it’s the same spot. How is your survey done? If it’s with a typical GPS, those values can be out by a good amount, depending on how many satellites were around at the time.
It could be that you’re trying a location that really does have a problem with the data. Can you give a street address or coordinates that I can try, to see what results I get?
I realized after posting that you may have been talking about elevation. There are a specific number of terrain values in each import. If you were to do an 18X import of the center of the area, then add imagery to get the surrounding 18X areas, that will give you more detailed terrain.
Hi Colin,
Yes my issue is about the Z information. As I said I wasn’t expecting it to be a perfect fit but 3m out? The X&Y are close enough. I imported at 17 or 18. Are any of the paid options any better? What other formats will Sketchup import for terrain data? GML 3.2, ASCII grid for DTM, GML 3.2 and Esri shapefile?
This is an old post, but you could see how this extension is doing, or do other searches in the forum about importing data:
With the built in Digital Globe and Nearmap, the terrain data comes from someone else, and it’s the same source for both tile image providers.
I think that it isn’t just being 18X, but also selecting the smallest area. I’m going to test that on a hilly area, to see what level of detail I get.
You need an area to check back against or to be able to check against OS data. My data is from a topological survey of the site. More than happy to send it to you together with location information.
The LIDAR takes many more readings than the surveyors are doing on the ground. Maybe the inaccuracy is in the smoothing and simplification to lower the poly count.
That gave me an idea. I was trying to subjectively see if I noticed differences in terrain detail when importing a large area at 18x, a smaller area, and a smallest area. Here’s a screenshot that shows the three of them, with hidden geometry turned on. The smallest area one does seem to have more faces for a given area.
yes but is the data anymore accurate? It may be subdividing the smaller grid into finer segments without improving the accuracy of the information
Yes it makes no difference in terms of the height accuracy. The small square is at level 18 and as small as possible.
This has been a recurring issue I don’t have a good, or at least quick and easy answer to. Years ago it was Google Earth data that SU brought in. I remember in the past getting by with just vertically scaling the imported terrain until it was close enough at the boundaries to be believable. Most recently, I tried two sources via PlaceMaker with higher hopes, and the terrain still wasn’t great. Just a couple days ago, I spent an unconscionable amount of time creating surrounding context terrain that would mesh perfectly with the A2 survey contours from my surveyor at the property edges. I have to believe and use the survey for my project property, but the surrounding terrain is just there to provide context, and only needs to be a good enough representation. The problem is merging the two at the boundaries; that in a nutshell is the issue, yes?
Actually I think it is satellite data. Our National Survey has made its aerial Lidar data public and compared to that the Digital Globe and before that, Google data is vastly inaccurate. Even the National Survey data is not accurate enough to place building levels, with points about 2 m apart. They are actually doing a new scan of the whole country at a resolution of something like 3 points/m². That would probably be quite usable.
If the levels are consistently off the problem might be with coordinate systems. Surveyors use local “flattened” map coordinates (like the ETRS-GK…/N2000 here in Finland) while the Google Maps/Digital Globe etc… use a very different spherical (lat, long…) system
Hi RT,
Yes that is the main issue in a nutshell. However, as in this case the area sits in a dip that has public paths around it so I have to design with privacy in mind. Accurate heights of the surrounding areas would help with this. I also need to know whether the surrounding areas affect the availability of light for plants which isn’t usually so much of an issue but the hills around this site are much taller than the downloaded terrain claims it to be.
I have spent all afternoon looking for a way to even check altitude at a point. The results were at best poor. I did find one discussion on the data used to create the height maps and it also said that the data is about to be updated to be more accurate. Trimble should put a warning on it so people are aware of the issue.
You’re working with SketchUp’s built in terrain import? Have you tried PlaceMaker? It might have better results for your case.
