holes in point grid at fine spatial resolution #535
-
Hi, I am enjoying exploring MintPy. Thanks again for sharing it. It sounds like an exciting time for SAR with ICEYE and Capella Space entering the space. Suppose I am trying to get the most resolution I can out of Sentinel. The ISCE
And then doing something very minimal with MintPy, after loading up the ISCE stack per documentation.
Creating a shapefile Here this is a toy example with only three snapshots. I see the same thing when looking at years of data with more thoughtful parameters (NESD, Snaphu, MintPy network modification etc). I wonder if this is because the SRTM DEM is only 30m? Or perhaps ISCE or MintPy isn't intended for such a small number of range and azimuth looks? I want to explore replacing ISCE with Fringe but have had issues isce-framework/fringe#41. Advice invited. |
Beta Was this translation helpful? Give feedback.
Replies: 3 comments
-
If you have a stack in radar geometry, you are much better of generating the point cloud from that rather than from geocoded results. Geocoding inherently will force it on to a 30m grid. If you are looking for resolution, the radar geometry lat/lon should be better. Coarser azimuth resolution will dominate and you will still see linear patterns with S1 data. |
Beta Was this translation helpful? Give feedback.
-
By default, mintpy is geocoding in roughly the same grid size as the input file in radar coordinates (resample.py#L327) using the nearest-neighbor interpolation. I second @piyushrpt recommendation on feeding the radar-coord files to |
Beta Was this translation helpful? Give feedback.
-
Thanks! I didn't strictly realise I could combine inputs and outputs in the It looks much nicer now. |
Beta Was this translation helpful? Give feedback.
If you have a stack in radar geometry, you are much better of generating the point cloud from that rather than from geocoded results. Geocoding inherently will force it on to a 30m grid. If you are looking for resolution, the radar geometry lat/lon should be better. Coarser azimuth resolution will dominate and you will still see linear patterns with S1 data.