The way EA models various RDB objects (tables, their indices and constraints, views, etc.) with UML - where exactly did it come from?

Is there any sort of standard or recommendations on how RDB and UML concepts should/could be mapped to each other? Or is it completely homegrown by Sparxians?
The reason I am asking this is to understand exactly what is EA's freedom in how this modeling could work (and then suggest possible changes to how it works now).
Thank you,
Yarik.