In this video tutorial we will use Sentinel 1 data to map urban footprints We will do that using both the amplitude and the phase of the signal We will use 2 Sentinel-1 images we will calculate the amplitude the calibrated amplitude for each image and will also calculate the interferometric coherence between the two images And we will use the amplitude and the coherence to mask out the urban areas So let's get started First we will open the two Sentinel-1 images So here we have the Sentinel-1 images in the product Explorer window We can see from the filename The acquisition dates, so one was acquired on the 2nd of January 2016 the other on the 26th of January If we look at the bands we can see that the images are complex We have the complex bands, and we also have the virtual intensity band The first thing we will do is to subset these images because we do not need the whole image So we go to radar Sentinel-1 TOPS Sentinel-1 TOPS split and Here in processing parameters Here we can zoom in to the footprint of the image And we can select the various different sub swaths and the various different bursts, so we want to select iw 1 and bursts 4 to 8 In the i/o parameters we can select our output folder And the output file name by default it adds an underscore split to the end of the original file name and we click on run I will do all the processing on one of the images and then the processing again on the second image And then finally we all merge them together So here we have split product The next thing we will do is to apply precise orbits So we go to radar apply orbit file we select the split product and here under processing parameters, we can select do not fail if new orbit file is not found The precise orbits will be downloaded automatically from the internet If for any reason it does not work, then the process will not stop So now we can select our output file name, and we can select run The precise orbits will give us better orbit information which will improve the geometric correction and the co-registration And the reason why we have to download this from the Internet is because when the product was created The precise orbits were not available So now we will do a calibration So notice how the The bands we have at a moment are the complex bands, and we have an intensity band Which is a virtual band created on the fly We will now Calibrate this image to create a band of Sigma naught backscatter So we go to radar radiometric calibrate And here we select the last the product with the precise orbits and In the processing parameters, we will leave everything as default so we will output a Sigma naught band We click on run The next process is to Perform a deburst So here we have the calibrated Sigma naught band if we double-click on that we can view that in a viewer the debursting will remove this space in between the various bursts It will stitch these together And it'll remove these no data values so we go to radar Sentinel-1 TOPS Deburst And we can leave everything as default notice how it creates an underscore deb for deburst Okay, that's finished it took 31 seconds, so here we have our debursted image Okay we no longer see those there is black lines in between the bursts The next thing we will do is to do a multi looking This image is in single look complex format The dimensions between the pixels in x and y are not the same So we will create square pixels by selecting radar Multi looking And in the processing parameters we will leave the default multi look factor to create square pixels Which in this case is 3 by range and 1 by azimuth And we will select run The multi looking will also reduce the speckle by some degree as it averages out some of the pixels in one dimension So here we have a multi looked image We double click on that you can see the result okay, so now we have square pixels Notice how the histogram here is not so easy to manipulate What we commonly do is to convert the band from linear to a logarithmic scale and that will create a histogram that's much easier to to work with It improves the visualization of the image, so here we will create a new virtual band Which is in logarithmic scale in decibels So if we double click on that here we have a histogram, that's much easier to work with The image is easier to visualize Finally what we'll do is to geometrically correct the image In fact we will do a terrain correction that uses also a DEM to correct for the terrain So we go to radar Geometric terrain correction Range Doppler Terrain Correction Here in the processing parameters We can leave everything as default so we will use the SRTM 3 arcsecond DEM which will be downloaded automatically We will keep the default pixel spacing and the default map projection we will check the input output parameters Then we will select run The software will now look for the relevant DEM tile that covers this area, and if it does not find it in the relevant folder, then it will download it automatically from the Internet In this case I already have the DEM tile so it does not need to download it So that's done. It was completed in 54 seconds We can now close this window and here we have our final terrain corrected product We will again create a virtual band in decibels And here we have our terrain corrected calibrated multi looked image in decibels So we will repeat all of these steps for the second image, which I'll not do here in this video tutorial once we have done that we'll have both of the images calibrated and terrain corrected What we'll do is to now create Another image, which is the coherence image so to do that I'll close everything in my product Explorer window So to do that I will open the two images That have already been split Already have precise orbits So here we have one And here we have the second To calculate the coherence first we need to co-register the two images and then we can estimate the coherence Once we have done that we will do TOPS deburst a multi looking and a terrain correction of the coherence image So first of all to do a co-registration We select radar Coregistration Sentinel-1 TOPS Coregistration, and we will select Sentinel-1 back geocoding Here we can select this icon to add the images that are opened We will leave everything as default What we will do here though is to change the filename to leave only the common parts That are relevant to both Images so we will remove the acquisition date and various other information and leave only the common parts Then we select run So the coregistration will resample one image onto the other This has to be done very precisely For proper calculation of the interferometric coherence The pixels from both images need to match to a very high precision Well below one pixel Here we have our coregistered stack we can now close this window The next step is to calculate the coherence so we go to radar Interferometric products coherence estimation So we take as input the coregistered stack And we leave everything is default in the processing parameters section And then we select ok we select run Here we have our coherence product we can now close this window. Let's take a look So if we double click on that it'll open in a viewer And here we have our coherence image So you can see that areas with high coherence are white which correspond mainly to built-up areas where there have not been so many random changes between the two image acquisitions Whereas low coherence you can see a lot of low coherence in surrounding areas We will now go through the same processing chain that we carried out for the amplitude images So we will do a deburst, a multi looking, and a terrain correction So first let's do the deburst so radar, Sentinel-1 TOPS TOPS deburst And here we select okay we select run Then we do a multi looking Again we leave the default Range looks and azimuth looks in order to produce square pixels Now we will do a terrain correction so we go to geometric terrain correction, Range Doppler terrain correction You select the multi looked product, and we leave everything is default and we click on run You can now close this window and here we have our final terrain corrected coherence image So we can open this final image And see how it looks What we now want to do is to use this coherence image as well as the amplitude images from the two From the two image acquisitions, and we're going to try and map the urban areas with these three information layers So we will keep only this final this final processed coherence image And we will close all of the intermediate images so we can do that by selecting the final Terrain corrected coherence right clicking and select close other products Now we will open the two processed Sigma naught images So here we go to The two terrain corrected Sigma naught images So here we have the two terrain corrected multi looked calibrated Sigma naught images And we have the coherence And what we will do is we'll merge them together into a stack So we go to radar coregistration stack tools, create stack And here we will add opened to add all of the products We will leave everything as default here And we click on and we select run So here we have a stack with the coherence and the Sigma naught backscatter of the two images What we do is we will create some additional bands Including the average of these two backscatter images and also the difference But we will do that with the images in decibel So here we select linear to from dB When we create two additional virtual bands in decibel We will now save these as Not just virtual bands, but we will save them to the file so here we select convert band Convert band to each of these two virtual bands, then we save the image This way when we process these images in decibel We're not always processing the bands on the fly we're actually processing the data. That's written to a file And it should speed up the processing So what we will do here is we'll create the average of these two and we'll also look at the difference between these two Bear in mind that these are now in logarithmic scale and when you do a difference in logarithms It's the equivalent of taking the logarithm of the ratio and vice versa So to do that we go to raster band maths And here at first we will select the mean the average backscatter And here we will deselect virtual to write the newly created band to file and here we select edit expression Here we will Add the two backscatter bands together Divide them by two So here we have the mean dB, and we select okay Now we will take the difference so we will go to again two raster band maths And here we will select we will create a new band which will call difference Again we will deselect virtual and here we take one image and subtract the other image from it And select okay So here we have These new bands and what we will now do is we will open an RGB composite With these new bands And we will select the three RGB channels We will select in red the coherence image In green the mean image and in blue the difference image And select ok Here we have An RGB composite with the coherence the mean and the difference Now how do we interpret this image? the areas which are in red have low backscatter and high coherence These could correspond to agriculture or bare soil areas The areas which are in yellow on the other hand have high backscatter and high coherence These areas could correspond to built-up areas What we will now do is will try mask out these built-up areas by creating a new mask layer So we do that by selecting Raster band maths and here we can select a new band, which we'll call urban footprint And we'll create a new expression where we type in Conditional expression The conditional expression we can select from here from the operators And what we will do is we'll put in some thresholds in the conditional expression Which will help us mask out the urban areas so first of all we will select condition on the average backscatter on the mean backscatter so here we'll say if mean backscatter is greater than is greater than minus 10 And We will create condition on the coherence Here there's a slight, bug it does not appear where the cursor is so we have to modify this a little bit and coherence is greater than 0.6 Let's just put a space in between these two Then the mask will equal 1 otherwise it will equal 0 So here we're going to mask out the the yellow areas And then we select okay These areas should correspond to urban areas And then we select ok and here we have our mask We can compare this with the RGB image By selecting windows window tile evenly Ensuring that the cursor and the zoom are synchronized and here we can compare the two images We can also compare the two images by overlaying one on top of the other By selecting layer manager selecting the plus icon Image of band next and here we select our mask And we select finish now we can change the select with this Mask image and change the transparency to compare the two images This mask may not be perfect you can play around with the thresholds to improve the result in this tutorial we used two Sentinel-1 images in single look complex format to mask out urban footprints I hope you enjoyed the tutorial. Thank you for watching
