Change detection and ratio

With the ratio-image I mean for example:

ratio image = (SAR image at very windy conditions)/(SAR image at very calm conditions)

In calm conditions water surfaces have very low backscatter values, and in windy conditions the backscatter can be quite high -> ratio images detect water-bodies quite nicely.


very nice! Thanks for sharing it! Is there a specific rate expected that we should obtain from this image ratio (sth between a and b) specifically for water? In other words, how we can recognize e.g., image ratio of 3 is for the water bodies?

Thanks again

i try to execute a change detection with l1 data.
my proc.graph is: productsetreader->create stack -> cross correlation-> changedetection ->write
But everytime it appears the failure: “ArrayIndexOutOfBoundsExeption”
I figure out that if i split the chain in 2 parts, like this: productsetreader->create stack -> cross correlation-> write1
and then i start a new graph wich uses the write1-output for the the secondgraph: read(from write1)->changedetection ->write it works.
why dont it work if i use the composite graph?
Btw: productsetreader uses a orbit applied,radiometric, geometric calibrated, despeckled L1 data.

Hello A5iveMM,
maybe the problem is that your co-registration process lacks the warp step. Try to append warp after cross correlation, it should work.

Hi all

I need to make an change detection with ALOS PALSAR data, but in the snap I don’t see the option to introduced two images!

The snap gives me this message, but I only see one source product!!!

Someone know, please?

Thanks for advance

you first need to bring them into one new dataset. You can achieve this with
Radar > Coregistration > Coregistration

The new dataset will have both amplitudes and you can select it in the change detection module.


Dear Gabriela,
you have to create a co-registered stack first with your two images. I hope this helps.


1 Like

Thanks very much ABraun and gabbro, I did the change detection

This is the results

1 Like

Hi all

the change detection modify the river course

Images before and after

Result with the river’s course different!!!


I don’t understand why the river’s course is moved upwards

This looks like the coregistration was very bad. See what the RGB of the stack looks like.
Try with a larger cross correlation window.

1 Like

lveci is right. The images are not aligned correctly. If a larger cross correlation window doesn’t work either try terrain correcting both first and then create a stack with coregister > create stack

1 Like

I go make again, taking your recommendation and the recommendation of lveci!!!


Could it be also that Gabriela needs to apply the orbit files before coregistering?

@gabbro: It could certainly improve the result, good point.

Hi people

I did the coregistration, but I used NEST this time… I don’t understand why SNAP cannot makes an good coresgitration for ALOS PALSAR images

Best regards

Now, the change detection with the stack in NEST

Change Detection in Colour by SNAP, ALOS PALSAR images (08/07/2007 and 23/08/2007)


this is interesting.
How could the differences in registration accuracy arise?

I have some question about change detection
first of all what is the best data format for change detection? slc or grd / IW or EW or WV or SM/ and what polarization?

second is there any tutorial or step by step processing guide for doing change detection?