1. Existing landcover categorization schemes probably number in the thousands, notable ones are the ESA's Corine and the USGS NLCD, aligned to the areas of interest from planet, continent, region, down to small water sheds - i.e. they are customized according to the domain of knowledge and the area of interest according to what features are of interest and identifiable. DLR's global URBAN TEP ( https://www.dlr.de/eoc/en/desktopdefault.aspx/tabid-11882/20871_read-53736 ) is binary, urban and not. Some eco-region landcover have hundreds, discerning individual tree species.
2. Landcover 'code' for satellite imagery spans nearly fifty years since the first Landsat mission, with categorization schemes emerging during that span and some becoming standardized - it would be an open question what benefit a new one would have, since existing one have extensive papers written have been validated against ground truth for a considerable time. ( https://directory.eoportal.org/documents/163813/238965/History.pdf )
3. 'Satellite images' encompasses nearly six decades of acquisition, "More than 150 Earth-observation satellites are currently in orbit, carrying sensors that measure different sections of the visible, infrared and microwave regions of the electromagnetic spectrum." ( https://www.americanscientist.org/article/fifty-years-of-earth-observation-satellites ). Each platform ( satellite ) can have multiple sensors, which in turn have many bands whose selection, permutations, and combinations are use to feed any classification code. The bands range from thermal to radar wavelengths.
4. The above means there are tens of thousands of repositories of code for doing landcover mapping. I know there are probably hundreds on Google Earth Engine alone, and even more in R and Python.
It would be perhaps useful for one to be as specific as possible. The Earth is a big place.