Raster footprint generation

[drnextgis]

rio-stac creates a geometry property using a raster extent. I found this article explaining how to generate convex hulls of valid raster data for use in STAC Item geometries to be very useful. I believe it would be beneficial to have this option incorporated into rio-stac.

[vincentsarago]

thanks for sharing @drnextgis

I believe I've already implemented some of this in https://github.com/developmentseed/rio-stac/blob/main/rio_stac/stac.py#L53-L73 but I couldn't (or didn't want) to go further down the optimization because:

• I didn't wanted to add shapely dependency
• I didn't want to read the data (at least for the geometry part, we read the data for the statistics but we usually use overview)

[drnextgis]

Fair enough. What would you recommend for a user who loves rio-stac and desires to have an accurate footprint as the geometry of an item? 😊

[vincentsarago]

I guess you'll need to create some custom code

import rasterio
from rio_stac import create_stac_item

with rasterio.open("my.tif") as src:
    item = create_stac_item(
            src, 
            input_datetime=input_datetime,
            extensions=extensions,
            collection=collection,
            collection_url=collection_url,
            properties=property,
            id=id,
            asset_name=asset_name,
            asset_href=asset_href,
            asset_media_type=asset_mediatype,
            with_proj=with_proj,
            with_raster=with_raster,
            with_eo=with_eo,
            raster_max_size=max_raster_size,
            geom_densify_pts=densify_geom,
            geom_precision=geom_precision,
    )
    item.geometry = get_precise_geo(src)
    

[vincentsarago]

The more I think about it the more I think that maybe we could add optional dependency (shapely) to allow this!

Like if the user request precise geometry we check first if shapely is available 🤷