I suggest starting a single topic regarding the definition of the problem, and namely what should be detected, where all the question should be asked (and hopefully answered) so we know where to look. I'll start with a first set of questions:
* Does changes without human intervention should be considered? I can think of the evolution of water (sudden change of water or tides) or crop fields. Should we only detect permanent changes?
* Should defects in the images be detected as a change (for example a saturation that block the visibility)?
* Should a parking lot, and more generically moving vehicles, be considered as changes?
* Where should the changes be located when an image is badly aligned with respect to the rest of the timeseries? At the location inside that image or at the position of where it should have been if the image would have been properly aligned?
* If a building is constructed and destroyed during the same month (so that it looked exactly like it was before the construction), should it still be considered as change at the monthly level?
* Should we also include the 1st of the next month in the detection of the current months (for example 2018-01-01 to 2018-02-01 included for the month of January 2018 or 2018-01-01 to 2018-01-31, or the last date available for this month for the sequence)?
Hi, hopefully the clarification below will help you.
* the changes without human intervention, should indeed be considered as a change. So, i.e., vegetation => water should be considered as change. Additionally, the human interventions should be considered as a change (regarding to your 3rd bullet point).
* no, the defects in the images should not be considered as a change.
* it should have been where the image would have been properly aligned.
* no, as you specified if the building is not changed in the monthly basis, it is not a monthly change.
For the last bullet point, sorry I am not sure I understand, could you make it more clear?