Problem with split the image

as I said, if the method to derive the binary images produces unreliable results, you cannot use them for change detection.

So it’s right but for another I’ve images done those steps,for 17-18 too.
I’ve done the steps two time but the outputs are the same

I have a question,
I’ve finished the all steps for four years(four images)
According to steps we’ll have three images :
I know that I’ve made an image with mean and difference bands
But the binary image how have been created exactly ?
is it possible that 17-18 image buildings will be more than 17-20 ?

In my opinion" 17-20 image buildings should be more than 17-18 image buildings"
Is that right???

Honestly, how can I know? I don’t know your exact steps, nor do I know the city.
You are responsible for your workflow and the study area yourself.

But yes, the growth rate from 17-20 (three years) should be larger than from 17-18 (one year).

so i’ve sent for you the images,any thing no matter

profasser i should type a thing for the place of question mark it cannt be empty , can it be ‘or’?

  1. img1 == 0 and img2 == 0 and img3 == 0
  2. img1 == 0 and img2 == 1 ?
    img2 == 0 and img3 == 1
  3. img2 == 0 and img1 == 1?
    img3 == 0 and img2 == 1
  4. img1 == 1 and img2 ==1 and img3 ==1

with OR you indicate that there was growth between one of both periods.
AND does not make sense to me here.

For general changes, I can extract the differences between pairwise images and put the results together,yes