The "GeoAI / Remote Sensing" category turns satellite imagery into vector answers inside a single flow — masking cloud, computing indices, comparing dates, and converting what's left into polygons.
Start from imagery
These stages read from the "STAC Scenes" reader, which searches satellite archives and feeds scenes into the flow. Pick the sensor profile matching your source, or choose the generic option when you're supplying your own imagery and quality mask. See Readers and Sources.
Build a detection pipeline
Most detection flows follow the same shape:
- "STAC Scenes" — search scenes for your area and date range.
- "QA / Cloud Mask" — drop cloudy and invalid pixels first, so they can't masquerade as signal.
- "Band Calculator" — compute the index that carries your signal, such as NDVI.
- "Temporal Composite" — collapse a stack of scenes into one stable raster per period.
- "Raster Threshold" or "Raster Difference" — isolate the pixels of interest, or the change between two dates.
- "Persistence Filter" — require the signal to recur before you believe it.
- "Polygonize" and "Regularize Footprints" — convert detections into clean vector features.
From there, summarize by area with "Zonal Statistics", or send the polygons through transform stages to a writer.
If you want detection running on a schedule with a review workflow instead of building the graph yourself, use the Remote Sensing pane.