Estimating buildings height


#1

How can I estimate buildings height from SLC images?
Though tough work, any practical idea is appreciated rather than just list of papers!


#2

That requires Persistent Scatterer Interferometry (PSI) - processing (not supported by SNAP) of a large co-registered stack (30+ scenes, can be done with SNAP). I’d say such processing is extremely involving requiring a very high level of experience and expertise. In other words, if you want to go this way you should probably seek out an expert who could do the processing for you as the learning-curve is really steep.


Does SNAP support PSI for surface deformation mapping?
#3

You can still roughly estimate building height information from a pairs of SLCs, since the output heights from InSAR (DEM extraction) are in fact height of the scattering and not the terrain/elevation. You should be careful though in the selection of the pair to avoid ones with high temporal separation (loss of coherence). Also you will need perpendicular baseline not to be very short in order to have some sensitivity to heights.
The resolution of the SAR system is important controlling the level of details to will get (single building, block of buildings etc.)
To reduce also the effect from atmosphere you might consider generating multiple outputs from different pairs.


#4

You are assuming that it is possible to unwrap the phase to the scatterer that is on top of a building. In general this will not be possible so PSI is needed IMO.


#5

I believe you can still get the heights at a pixel level even for high resolution sensors, but if you need info for single buildings then you should go for PSI.


#6

@mengdahl,@mfoumelis Thnak you very much for letting me know about PSI. If it estimates bettwe than just pairs of SLC images, I thought why not to try it. So, interested to know what are the basic idea/method behind it after geocoding several pairs?


#7

30+ scenes are needed for PSI. See section 4.3 in this ESA report for a brief introduction:

I also found this thesis online which might be useful for you:


#8

Hi Cheng,
Have you succed to measure the height of Buildings. Please share your experience, I badly Need this.


#9

Hi mfoumelis,
I am trying to calculate building heights from pairs of SLC. After unwrapping and snaphu importing, I tried to run ‘Phase to Height’, but it didn’t Show any result. what could be the reason and am I right?


#10

Mitun, AFAIK ‘Phase to Height’ doesn’t work. You can try ‘Phase to Elevation’ instead. You would see Elevation band (Elevations from DTM you’ve used (SRTM or something)) and Elevation_VH (elevations on your DEM).


#11

I don’t think building height extraction will work as the phase will most probably be impossible to unwrap.


#12

I have found the following results: after phase to elevation, I have added elevation band (SRTM 1arcs), created linear/db for both and subtracted SRTM elevation from my output. Here it looks like:


Also, I have filtered with mean. But the height difference is pretty weird which cannot be count as Building Height. But the built-up area is clearly visible. Here is the SRTM elevation:


And the study area is:

It would be really helpful if you help me to find a way to get the building heights. Thanks again


#13

I agree with mengdahl: Buildings are relatively small targets which are probably not fully covered by SAR phase at medium resolution. You could, as shown above, get some mean elevations or heights, but discriminations of buildings from ground surface is nearly impossible without LiDAR data unless you use PSI techniques on really high resolution data.