StaMPS-Visualizer, SNAP-StaMPS Workflow

weiweian · March 24, 2021, 3:50pm

Hello @thho , Before I export, I see that lonlat has only four decimal places. After I export, I see that lonlat has five decimal places in the csv file, which is very strange to me. In addition, I don’t know how to modify the number of decimal places displayed in the matlab workspace, so I can view the above numbers by copying them into the text.

thho · March 24, 2021, 4:13pm

hmm I see, too bad that I do not have access to Matlab and some playground data. Without I am not able to figure out what is going on in detail and to me it sounds like that there is some thing going on with just one or two objects which have to be looked at closely…

weiweian · March 24, 2021, 4:30pm

Yes, I think so too, if I solve the problem I will come back to give the answer, anyway, have agood day Thanks again.

despina98 · March 26, 2021, 1:22pm

hello,
This message appears when I run stamps(7,7)

, but the processing continues. what is the reason?

suribabu · March 29, 2021, 4:50pm

Dear Gijs,
I just processed 52 images using STAMPS, and I get LOS displacement which was exactly matching with GPS site placed over my AOI.

As per your studies we need both ascending and descending to compare the LOS with vertical displacements. But I have only descending orbit data only.
How can I compare this results ? If you have any python code for single orbit data(converts LOS to vertical ) please share it.

ABraun · March 29, 2021, 6:07pm

what’s wrong with his answer here? Documentation on SNAP-StaMPS workflow

suribabu · March 30, 2021, 2:14am

Nothing wrong Dear ABraun, In this below link vertical_displacement_calculator.py code is present. I want to check it this python wrapper to descending orbit data set.

Thank you.

gjvanleeuwen/SNAP-StaMPS-Visualizing: Documentation on the bachelor thesis: AN ASSESMENT OF SENTINEL-1 RADAR IMAGERY TIME-SERIES ANAYLIS FOR MONITORING SEASONAL SURFACE SUBSIDENCE DUE TO PERMAFROST DEGRADATION AFTER A TUNDRA FIRE (github.com)

weiweian · April 1, 2021, 11:37am

I have used the csv file obtained by StaMPS-Vsualizer. How does StaMPS calculate the average deformation rate based on the settlement?

thho · April 3, 2021, 10:25am

To which column do you refer to?

weiweian · April 3, 2021, 10:57am

The third column，mean velocity

thho · April 3, 2021, 1:02pm

As far as I understand it, over the entire time period of your stack (lets say its 2 years for example), the velocity is calculated by looking at the absolute difference between the first and last point (lets say for one point that is 4mm) then you have a 2mm/yr velocity.

However, there might be other ways to calculate that. e.g. calculating the velocity for the period 1st image to prime and a second velocity for prime to last image and then the mean of both results. But as far as I remember (I looked into this in 2017/2018) It was done as described in the first option

weiweian · April 3, 2021, 1:37pm

Yes sir, you made a correct point, but there are still some differences between the average speed calculated by the first method and the value directly derived from StaMPS. I don’t know where these differences come from. Do you have any good suggestions for this, thank you！

thho · April 4, 2021, 8:58am

No other than the two options I mentioned. But I looked into the MATLAB code of StaMPS and localized the lines where mean velocity is calculated:

some objects are prepraed before but the main part is from line 196 to 204:

github.com

dbekaert/StaMPS/blob/master/matlab/ps_mean_v.m#L196


ifg_cov=ifg_cov(unwrap_ifg_index,unwrap_ifg_index);
if rank(ifg_cov)<size(ifg_cov,1)
    error('Interferograms included for which there is no unwrapped solution')
end
if use_small_baselines==0
    day=ps.day(unwrap_ifg_index)-ps.master_day;
else 
    day=ps.day(ps.ifgday_ix(unwrap_ifg_index,2))-ps.day(ps.ifgday_ix(unwrap_ifg_index,1));
end
N=length(day);
ph_uw=ph_uw-repmat(nanmean(ph_uw(ref_ps,:),1),ps.n_ps,1);
lambda=getparm('lambda');
ph_uw=double(ph_uw/4/pi*lambda*1000)'; 
G=[ones(N,1),double(day)/365.25] ;
% computing mean velocity
mean_v=lscov(G,ph_uw,ifg_cov)';
mean_v=mean_v(:,2);

As you can see, it is not that straight forward as described, which explaines the little differences. If you like read the script closely and when you post a summary here about how exactly the mean velocity is calculated we would all be very thankful

weiweian · April 4, 2021, 11:29am

Sir, I did some tests with the exported data according to the code in ps_mean_v.m, and found that the results obtained are satisfactory！

The data in ph_uw represents the cumulative deformation of each imaging time,As shown

G represents the matrix of imaging time：

the most important is：mean_v=lscov(G,ph_uw,ifg_cov)’;Since I don’t know what ifg_cov represents, So I use mean_v=lscov(G,ph_uw)’; command to test.

The results show that the difference between the calculated average deformation rate and the result derived by StaMPS is less than 1

Finally, attach a description of the lscov function in matlab:

thho · April 4, 2021, 12:28pm

@weiweian Thanks for sharing your insights! That will help others to understand this value better. The question about the precise calculation of mean velocity came already up once or twice.

SteffanDavies · April 19, 2021, 11:40am

Dear thho, do you have any idea how I can force 0 to be the mid-point of the colour ramp?

ABraun · April 19, 2021, 6:21pm

you have to set an area as reference with either

ref_centre_lonlat: [0 0]
ref_radius: Inf

the area you select here will be the zero-displacement reference, so depending on which point you select, this will affect the legend. If nothing is defined, StaMPS takes the average displacement of the scene, but this can differ from the true zero-displacement depending on the size of the scene and the displacement pattern.

thho · April 20, 2021, 6:57am

Hi @SteffanDavies,

if you want to do this in StaMPS-Visualizer, this is tricky I think, here is how it works in general:

the colorramp is defined by the min and max of your displacement values which you have exported from StaMPS in Matlab.
Since the distribution can have many characteristics like all positive or negative, 90% over 0 and 10% under 0 (e.g.) the color ramp it is not designed to have a static center but to adjust to every new dataset and give a high resolution of the displacement values aka make a min-max stretch very time
the line where this is specified is:

github.com

thho/StaMPS_Visualizer/blob/ad9b4edd4d2c670f100e072602e13a16544308da/server.R#L51


              dates.date=dates.date,
              ref.points=ref.points,
              ps.loc=ps.loc))
}
# render StaMPS points on interactive map
render.stamps <- function(pointsize = 5, setview = TRUE){
  disp <- stamps$ps.loc$disp
  colnum <- colorNumeric(colramp,
                         domain = c(min(disp),
                                    max(disp)))
  if(setview == TRUE){
    # show PS on map
    leafletProxy("map", data = stamps$ps.loc) %>%
      clearGroup("Measurement Points") %>%
      # add PS circles with color ramp
      addCircleMarkers(stamps$ps.loc$lon,
                       stamps$ps.loc$lat,
                       layerId = stamps$ps.loc$uid,
                       radius=pointsize,

here we take the pre defined matlab like color ramp object colramp and force the domain to be the min max of the displacement value (which can be seen as a min max stretch)

The new object colnum is later used here to set the pal parameter in the map legend:

github.com

thho/StaMPS_Visualizer/blob/ad9b4edd4d2c670f100e072602e13a16544308da/server.R#L69


                     stamps$ps.loc$lat,
                     layerId = stamps$ps.loc$uid,
                     radius=pointsize,
                     fillColor = colramp[as.numeric(cut(disp, breaks = 10))],
                     fillOpacity = 0.8, stroke = F,
                     group = "Measurement Points") %>%
    setView(lng = median(stamps$ps.loc$lon),#+0.02,
            #TODO: check if control panel toggle is True if, shift lng 0.02
            lat = median(stamps$ps.loc$lat),
            zoom = 15)%>%
    addLegend("bottomleft", pal = colnum,
              values = disp,
              title = "mm/year",
              layerId = "uid",
              opacity = 1)
}else{
  # show PS on map
  leafletProxy("map", data = stamps$ps.loc) %>%
    clearGroup("Measurement Points") %>%
    # add PS circles with color ramp
    addCircleMarkers(stamps$ps.loc$lon,

however, the points color is defined directly via the displacement values:

github.com

thho/StaMPS_Visualizer/blob/ad9b4edd4d2c670f100e072602e13a16544308da/server.R#L62


                                  max(disp)))
if(setview == TRUE){
  # show PS on map
  leafletProxy("map", data = stamps$ps.loc) %>%
    clearGroup("Measurement Points") %>%
    # add PS circles with color ramp
    addCircleMarkers(stamps$ps.loc$lon,
                     stamps$ps.loc$lat,
                     layerId = stamps$ps.loc$uid,
                     radius=pointsize,
                     fillColor = colramp[as.numeric(cut(disp, breaks = 10))],
                     fillOpacity = 0.8, stroke = F,
                     group = "Measurement Points") %>%
    setView(lng = median(stamps$ps.loc$lon),#+0.02,
            #TODO: check if control panel toggle is True if, shift lng 0.02
            lat = median(stamps$ps.loc$lat),
            zoom = 15)%>%
    addLegend("bottomleft", pal = colnum,
              values = disp,
              title = "mm/year",
              layerId = "uid",

So what you have to do, in order to center around 0 individually

create a custom color ramp by stacking two color ramps
- first ramp is e.g. from +10 to 0; blue to white
- second ramp is e.g. from 0 to -90; white to red
- stack both color ramps to gain a +10 - 0 - -90; blue - white - red gradient color ramp
then, this new color ramp has to be implemented to the two cases (legend color and point color)…how to do this exactly, I have to test myself…tbh

here is a code example that can be used, it is somewhat tricky I think, at the moment I do not have the time to implement it, but I will keep it in mind for the future

Giova · April 21, 2021, 7:22am

Hi @ABraun,

if I set ref_centre_lonlat and ref_radius so take a PS as zero-displacement reference (a stable rocky outcrop, for example), when I visualize the graph of this point on StaMPS-Visualizer it will be displayed as a horizontal line at zero (translation of every point data to zero) or only the mean velocity of this PS will be set to zero, or it will affect only the legend but not the data of displacement? Will the velocities of all others PS be affected by the setting of reference point?

Thanks.

ABraun · April 21, 2021, 8:51am

yes, the entire range of values will be shifted so that the area you selected as reference is zero. If this value also represents most of the unchanged areas, the legend will also be zero for this color.