Export DEM resolution

[bvpham]

Hi,

I built a 3D surface using aerial images. When I choose to export the DEM, the proposed resolution is 0.95 meter. I wonder if this value is the average distance between two points in the DEM? I choosed Medium setting when I built the geometry of the surface. I wonder if I choose High or Ultra High, the resolution will be improved ? (I know that I can choose other value besides the proposed value but I think that the software will interpolate the points if I choose a smaller value (e.g. 0.1 meter)).

------------------------------------------------
I rebuilt the surface with "High detail" setting and obtain a resolution of 0.75 m/pixel. With this result, I think that the proposed resolution when exporting a DEM is the average resolution of the built surface.

When I change from the "Medium detail" setting to "High detail" setting, I reduced the number of the images from 80 to 25 (my laptop crashed with 80 images in "High detail").

[Alexey Pasumansky]

Hello bvpham,

The best DEM resolution is defined by the effective model resolution. Which is determined by quality used for geometry reconstruction. The higher is the quality the more details you'll see in DEM exported.

Usually DEM proposed resolution by PhotoScan is slightly greater than max effective one. But is you set more detailed resolution than it could be acquired using the current model, it would be interpolated.

[Ludwig Hilger]

Hello everybody,

I am posting in this topic because my question is somehow related...
I am new to PhotoScan and I love it! I have completed all the tutorials and now want to test on my own data. I have created a mesh (decimated to 200000 faces) from 39 drone images and now want to export a 1 m dem (horizontal pixel size) with a vertical resolution of 10 cm.
I have always been wondering what the "H" and the "V" mean in the export DEM dialogue (is it the horizontal and vertical resolution?). There is also  no mentioning of the unit  these two parameters have (I guess meters?). I am also wondering what the value is you can enter under "Metres..." in the same dialogue (it changes if you change "H" and/or "V").
I have entered 1 in the "Metres option" and now my Computer seems to compute itself to death, which can't be because the area is rather small.

So I am wondering what I have to enter if I want to get the DEM with the defined resolutions mentioned above?

Other question:
a) Is there a rule when to use the "split in blocks" option? Too many blocks seem to slow down the process as well...
b) I get fantasy values for location and height although the relative values are correct (I can see the terrain and everthing). The numbers in the Ground control panel are correct as well.

Any help would be appreciated very much!!!!

regards,
Ludwig

[bvpham]

Hi Ludwig,

It's been a while that I do not use the software (we don't have many chance to fly a UAV around human area ). So, try to check if my answers are still ok.

As I remember correctly, if you have geo-referenced your image then the metric unit should be meter (if you provide ground control position in meter).

If you want to export a DEM, chose File then Export DEM. Then in the projection plane, choose the one that suits you. Normally, I choose TopXY. Concerning H and V, if I recall correctly they are actually Horizontal and Vertical but not in the sense that you suppose them to be ^_^

Actually, H and V here correspond to x and y axis in a x, y, z coordiate system where z indicates the height (I'm not so sure about this but I had this impression as I also wanted to have a vertical resolution like you). Normally, I have the same value for H and V. If you really want to have a fine (small value) vertical resolution, you should export the DEM as it highest quality (default value when you open the DEM export window). Then you can use other tool to down-sample the DEM like geo_translate utility (free linux and windows tool).

About the processing speed, we can think about days if you have a lots of pictures (more than 15 is a lot ) and if your computer is not so strong (mine have 12 Gb of RAM and 3.2 GhZ Quad core i7 but it's still only ok in doing its job).


1. I do not fully understand you "block" question. I used block features before as the area we try to scan is quite large. Actually, using the block could somehow reduce the precision of your map but you could save a lot of time. I don't have much experience in this one neither.

2. How could you check if precision of your ground control point? By exporting the DEM and comparing the point's coordinate with the entered values? I had this issue before. Actually, there are somekind of bend (the 3D map is bend), rotation error (the 3D map is rotated a few degress) as there are also error in the ground control coordinate. The rule is to have as many as ground control as you can and as spreading as possible. When we do our job, we put large, easy-recognible targets on the ground before launching the UAV. And we try to use the best Gps that we can afford (1 cm accuracy with respect to the base station).

Hope this help you somehow. Cheers.

[Ludwig Hilger]

Dear bvpham,

thank you very much for the fast reply! Ok, I will just try to not change the settings of H and V which should give me the highest resolution. And resample afterwards to a lower resolution. I could solve the problem with the georeferencing. It just turned out that my SAGA GIS could not directly read the tif exported by the file correctly. I had to import and export it into/from ArcGIS first, than it worked. Now, it's position looks ok. We are also using targets in the field we are measuring with the DGPS. I guess it is best to measure them in WGS 84 (EPSG: 4979 <-?) which the UAV is using as well. I had measured the GCPs in epsg 25832 and I have some conversion problems now.
And yes, I will definitly talk to my professor for a suitable machine, my 8GB RAM and 2.2 Ghz are not enough...

thank you again,

Ludwig