Using change vector analysis with two S2 Images

snap
#1

Hello, It’s the first time im using the SNAP program and the S2 toolbox.
I want to do Change detection, most specificially CVA with multispectral bands. I searched everywhere in the literature and I understand the theory behind it but I can’t put in pratic.

Im aware that I have to use band maths in order to calculate the vectors and then it’s possible to see the diference obtained in those products. Is there any recipe available in order to do so?

Thank you

#2

As far as I know, there is not CVA functionality in SNAP. There are a few things you could do to have the two bands of interest collocated in the same product, you could define a vector or ROI in SNAP and use it to compare both bands in a scatter plot, but you cannot derive magnitude and angle maps within SNAP.

#3

Thanks a lot! Maybe ENVI has that functionality?

#4

these slides provide suggesions on how to implement the Change Vetor analysis based on four selected bands in the Band Maths:
Sutton (2012) Change vector analysis

#5

Thank you!

I’ve used two images (june and september 2018) in order to perform CVA using various Spectral Indices.

However I’m having doubts to pick a threshold to define change/no change: i understand It is used a standard deviation but I don’t know how to perform and to apply in SNAP

The image and the histogram obtained: Screenshot - 3fbc644cfa728b0286edde4f1bd010f4 - Gyazo

Can you help me?

#6

you can create a scatter plot and use the “select mask pixels” to select parts in the scatter plot which are then colored red in the raster map. This might help you identifying those pixels which are responsible for change. Example: Scatter plot regions to image space and viceversa

#7

Thank you! I’ve been searching for that for some days, thats exactly what I wanted!