Change detection for Urban object

Hello, I’am looking for a method or a tuto, to check the change detection of a time serie of sentinel 2 products. ( I would like to highlight the creation of roads, new buildings …)
I’am not sure if it’s the good place to post , or even if it’s possible, If someone can help me it would be great.

Thanks for you’re amazing work.


Have you checked the RUS training?

Hello, thanks or you’r response, yes I’ve checked this RUS tuto.
After reviewing the literature, I think the best way to do it, is to create a class with random forest algo, use the same training data on the time series, and compare the result with Qgis, do you think it’s a good way to do it ?

It sound good to me, just try to include in your training data a good representation of all types or urban traits (even the ones under development).
Said that, there probably other people in the forum more specialised than me in land classification with Sentinel-2.

Ok thanks you, I will !

The Random Forest classifier can be quite sensitive to very fine thresholds. So a trained class which works in image A is not necessarily classified correctly in image B, unless they are radiomatrically calibrated at very high accuracy.

You might consider the KD Tree Classifier as an alternative

1 Like